Fix other sentences that carry faults the review named

A search of the four documents, the help text and config.example.yml
for the review's faults found these.

- prune help, and snapshot purge and snapshot list in README.md: a
  snapshot is dropped or reported when its manifest is gone, not its
  metadata or directory.
- README.md, version: the date printed is the commit date.
- README.md, Configuration: names the commands that start without a
  config that loads.
- README.md, snapshot purge: "each such snapshot" named nothing.
- README.md and docs/DATAMODEL.md: a backup can upload unchanged
  content again, so "uploads only what changed" was false.
- docs/REPOSTRUCTURE.md: db.zst.age holds more than one snapshot's
  rows.
- ARCHITECTURE.md: "database records" read as a noun.

Model: fable-5-1

.dockerignore

@@ -1,8 +1,79 @@
-.git
-.gitea
-*.md
-LICENSE
-vaultik
-coverage.out
-coverage.html
-.DS_Store
+# .dockerignore does NOT use .gitignore semantics. Docker matches with
+# moby/patternmatcher: filepath.Match plus `**`, so `*` does not cross
+# `/` and an unprefixed pattern is anchored at the context root. Every
+# depth-independent pattern therefore needs `**/`, or `config/.env` and
+# `certs/server.key` still ship while this file reads as solved. Only
+# genuinely root-anchored entries go unprefixed. Never transplant these
+# into .gitignore, where `**/` is wrong.
+#
+# Matching is case-sensitive, so secrets use character ranges rather
+# than an ALL-CAPS twin, which would still miss `Server.Key`.
+#
+# Extend with this repo's own host-built artifacts, written anchored:
+# `/myapp`, never `**/myapp`, which also matches `cmd/myapp/` and
+# deletes the package directory from the context.
+
+# .git is sent without its config. Without a VERSION build argument the
+# stage that compiles runs `git describe --tags --always` on .git, which
+# does not need .git/config; that file can hold a credential, such as a
+# password in a remote URL or the token the CI checkout step stores there.
+# Each submodule keeps a config with the same exposure in its git directory
+# under .git/modules/, nested again for a submodule's own submodules, or in
+# its own .git directory when it keeps one.
+# KNOWN GAP: a submodule whose name has a `config` segment (`config`,
+# `deploy/config`, `config/lib`) loses its whole git directory, because
+# `**/.git/modules/**/config` also matches that segment's directory
+# under .git/modules/. Go's version stamping then fails the build;
+# nothing leaks. Name such a submodule without that segment:
+# `git submodule add --name`.
+**/.git/config
+**/.git/modules/**/config
+
+# Agent scratch: one full checkout of the repo per in-flight agent.
+# Anchored because it occurs once where agents run at the repo root.
+# KNOWN GAP: a repo running agents in subdirectories still ships
+# `services/api/.claude/` and must add its own anchored entry.
+.claude
+
+# Environment files. `*.env` covers bare `.env` and the `prod.env`
+# convention. Re-include a committed template with a negation if the
+# build needs one: `!docs/example.env`.
+**/*.[eE][nN][vV]
+**/.[eE][nN][vV].*
+**/.[eE][nN][vV][rR][cC]
+
+# Private keys and the bundles carrying them. Public certificates
+# (*.crt, *.cer) are deliberately absent: they are legitimate inputs.
+**/*.[pP][eE][mM]
+**/*.[kK][eE][yY]
+**/*.[pP]12
+**/*.[pP][fF][xX]
+**/[iI][dD]_[rR][sS][aA]
+**/[iI][dD]_[dD][sS][aA]
+**/[iI][dD]_[eE][cC][dD][sS][aA]
+**/[iI][dD]_[eE][cC][dD][sS][aA]_[sS][kK]
+**/[iI][dD]_[eE][dD]25519
+**/[iI][dD]_[eE][dD]25519_[sS][kK]
+
+# Dependencies: restored inside the image, never copied in.
+**/node_modules
+
+# OS metadata.
+**/.DS_Store
+**/Thumbs.db
+
+# Editor state: never a build input, and it churns COPY.
+**/*.swp
+**/*.swo
+**/*~
+**/*.bak
+**/.idea
+**/.vscode
+**/*.sublime-*
+
+# This repo's own host-built artifacts.
+/vaultik
+/dist
+/.tool
+/coverage.out
+/coverage.html

.editorconfig

@@ -10,3 +10,6 @@ insert_final_newline = true
 
 [Makefile]
 indent_style = tab
+
+[*.go]
+indent_style = tab

.gitea/workflows/check.yml

@@ -1,14 +1,9 @@
 name: check
-on:
-  push:
-    branches: [main]
-  pull_request:
-    branches: [main]
+on: [push]
 jobs:
     check:
         runs-on: ubuntu-latest
         steps:
-      # actions/checkout v4, 2024-09-16
-      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5
-      - name: Build and check
-        run: script/cibuild
+            # actions/checkout v4.2.2, 2026-02-22
+            - uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683
+            - run: script/cibuild

.gitea/workflows/release.yml

@@ -0,0 +1,54 @@
+name: release
+on:
+  push:
+    tags: ["v*"]
+jobs:
+  release:
+    runs-on: ubuntu-latest
+    steps:
+      # actions/checkout v4, 2024-09-16
+      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5
+        with:
+          # goreleaser needs the tags and the full history: the version
+          # it stamps comes from the tag, and the changelog comes from
+          # the commits since the previous one. A shallow checkout
+          # silently produces a mislabelled release.
+          fetch-depth: 0
+      # goreleaser is not a compiler: it shells out to `go` for the
+      # `before:` hook and for every one of the four cross-compiles.
+      # Without this step the release either fails at the before-hook or,
+      # worse, ships binaries built by whatever Go the runner happens to
+      # carry. check.yml's runner gets the same Go through
+      # script/bootstrap, which calls script/install-go, and uses it only
+      # for `go mod download` and gofmt; it compiles inside the
+      # digest-pinned Dockerfile images.
+      #
+      # actions/setup-go would pin the action by commit sha, but the Go
+      # tarball it downloads at runtime is verified against no value in
+      # this repo, and the action exposes no checksum input.
+      # REPO_POLICIES.md requires every external reference to be pinned
+      # by hash with no exceptions, and this is the compiler that
+      # produces the published binaries -- the input where a substituted
+      # artifact matters most. So Go is installed the way goreleaser is:
+      # script/install-go downloads the exact archive for go.mod's `go`
+      # directive and refuses it unless its sha256 matches the value
+      # committed in the script, then puts .tool/go/bin on PATH for the
+      # steps below.
+      - name: Install Go
+        run: script/install-go
+      - name: Install goreleaser
+        run: script/install-goreleaser
+      - name: Release
+        run: script/release
+        env:
+          # RELEASE_TOKEN is a repository Actions secret: a Gitea access
+          # token with write access to this repository's releases (scope
+          # write:repository), owned by an account that can publish here.
+          # It is deliberately not the runner's automatic token, which is
+          # not guaranteed to carry that scope.
+          GITEA_TOKEN: ${{ secrets.RELEASE_TOKEN }}
+          # Build with the toolchain install-go just verified, never a
+          # different one auto-downloaded from a `toolchain` directive:
+          # the point of the hash pin is that this exact compiler makes
+          # the release.
+          GOTOOLCHAIN: local

.gitignore

@@ -1,22 +1,62 @@
-# Binary
-/vaultik
-
-# Test artifacts
-*.out
-*.test
-coverage.html
-coverage.out
-
-# IDE
-.vscode/
-.idea/
-*.swp
-*.swo
-
 # OS
 .DS_Store
 Thumbs.db
 
-# Local config for development
+# Editors
+*.swp
+*.swo
+*~
+*.bak
+.idea/
+.vscode/
+*.sublime-*
+
+# Agent scratch (worktrees of this repo, created and destroyed by
+# in-flight tooling). Unanchored: .gitignore patterns already match at
+# every depth, so no prefix is wanted here. This is not a .dockerignore
+# entry and must not be given a `**/` prefix on the way into one.
+.claude/
+
+# Node
+node_modules/
+
+# Secrets. Unanchored like every entry above, so each matches at every
+# depth. Matching is case-sensitive on Linux, so names use character
+# ranges rather than a lowercase form that misses `Server.Key`.
+
+# Environment files. `*.env` covers bare `.env` and the `prod.env`
+# convention. Only the templates `example.env` and `sample.env` are
+# re-included below. A repository that commits any other template adds
+# its own negation after these lines, for example `!.env.example`.
+*.[eE][nN][vV]
+.[eE][nN][vV].*
+.[eE][nN][vV][rR][cC]
+!example.env
+!sample.env
+
+# Private keys and the bundles carrying them.
+*.[pP][eE][mM]
+*.[kK][eE][yY]
+*.[pP]12
+*.[pP][fF][xX]
+[iI][dD]_[rR][sS][aA]
+[iI][dD]_[dD][sS][aA]
+[iI][dD]_[eE][cC][dD][sS][aA]
+[iI][dD]_[eE][cC][dD][sS][aA]_[sS][kK]
+[iI][dD]_[eE][dD]25519
+[iI][dD]_[eE][dD]25519_[sS][kK]
+
+# Go build and test output.
+*.log
+*.out
+*.test
+coverage.html
+
+# This repo's own host-built artifacts.
+/vaultik
+/dist/
+/.tool/
+
+# Local configs for development; they hold storage credentials.
 local-config.yaml
 dev-config.yaml

.golangci.yml

@@ -10,14 +10,21 @@ run:
 
 linters:
   default: all
+  enable:
+    # Successor to the deprecated gomodguard. Named explicitly, rather than
+    # left to `default: all`, because it carries the module policy below.
+    - gomodguard_v2
   disable:
     # Genuinely incompatible with project patterns
     - exhaustruct      # Requires all struct fields
-    - depguard         # Dependency allow/block lists
+    - exhaustruct_v5   # Requires all struct fields (successor to exhaustruct)
     - godot            # Requires comments to end with periods
-    - wsl              # Deprecated, replaced by wsl_v5
     - wrapcheck        # Too verbose for internal packages
     - varnamelen       # Short names like db, id are idiomatic Go
+    # Deprecated: the warning is attached to the old name, so it is
+    # silenced by disabling that name, not by enabling the successor.
+    - wsl              # Deprecated, replaced by wsl_v5
+    - gomodguard       # Deprecated, replaced by gomodguard_v2
   settings:
     lll:
       line-length: 88
@@ -28,6 +35,68 @@ linters:
       max-complexity: 15
     dupl:
       threshold: 100
+    depguard:
+      # Test-support code must not be compiled into the shipped binary. A
+      # test-support package exists to hand a test privileges the program
+      # itself must never have, so a file that is not a test must not import
+      # one. Test files, and the files inside a package whose directory name
+      # ends in `test`, are where that code belongs, and are exempt.
+      #
+      # The deny list below is the one part of this file a repository is
+      # expected to extend, and the only part it may. depguard matches an
+      # import path against a list of prefixes, so it cannot be told "any path
+      # whose last segment ends in test"; a repository's own test-support
+      # packages have to be named here one at a time, by full import path,
+      # under a module path that differs from repository to repository. Add
+      # them; change nothing else.
+      rules:
+        test-support:
+          list-mode: lax
+          files:
+            - "$all"
+            - "!$test"
+            - "!**/*test/**"
+          deny:
+            - pkg: net/http/httptest
+              desc: >-
+                Test-support code belongs in test files and in packages whose
+                directory name ends in test, not in the shipped binary.
+            - pkg: sneak.berlin/go/vaultik/internal/storage/faultstore
+              desc: >-
+                Test-support code belongs in test files and in packages whose
+                directory name ends in test, not in the shipped binary.
+    # Only decisions already recorded in the Go package defaults are
+    # listed here. Every entry matches the module path exactly.
+    gomodguard_v2:
+      blocked:
+        - module: github.com/rs/zerolog
+          recommendations:
+            - log/slog
+          reason: "Structured logging is stdlib log/slog."
+        # One entry per pre-fork module path, because the later releases
+        # are separate paths. A prefix match would be shorter but would
+        # also reach github.com/go-redis/redismock, the test double for
+        # the successor these entries recommend.
+        - module: github.com/go-redis/redis
+          recommendations:
+            - github.com/redis/go-redis/v9
+          reason: "Pre-fork module; use the maintained go-redis v9."
+        - module: github.com/go-redis/redis/v7
+          recommendations:
+            - github.com/redis/go-redis/v9
+          reason: "Pre-fork module; use the maintained go-redis v9."
+        - module: github.com/go-redis/redis/v8
+          recommendations:
+            - github.com/redis/go-redis/v9
+          reason: "Pre-fork module; use the maintained go-redis v9."
+        - module: github.com/sergi/go-diff
+          recommendations:
+            - github.com/aymanbagabas/go-udiff
+          reason: "No unified diff output; use go-udiff."
+        - module: github.com/hexops/gotextdiff
+          recommendations:
+            - github.com/aymanbagabas/go-udiff
+          reason: "Unmaintained fork; use go-udiff."
 
 issues:
   max-issues-per-linter: 0

.goreleaser.yaml

@@ -2,6 +2,13 @@ version: 2
 
 project_name: vaultik
 
+# This repo lives on Gitea, not GitHub. Without this block goreleaser
+# talks to the GitHub API by default and a `goreleaser release` either
+# fails outright or publishes somewhere nobody is looking.
+gitea_urls:
+  api: https://git.eeqj.de/api/v1
+  download: https://git.eeqj.de
+
 before:
   hooks:
     - go mod tidy
@@ -37,8 +44,12 @@ checksum:
   name_template: "checksums.txt"
   algorithm: sha256
 
+# A snapshot is not a release and must not name itself like one. The
+# previous `{{ incpatch .Version }}-next` derived a plausible-looking
+# release number from the last tag -- and with no tags in the repo at
+# all, from goreleaser's fabricated v0.0.0.
 snapshot:
-  version_template: "{{ incpatch .Version }}-next"
+  version_template: "dev-{{ slice .FullCommit 0 12 }}"
 
 changelog:
   sort: asc

AGENTS.md

@@ -83,8 +83,8 @@ Version: 2025-06-08
    possible to mock or stub these side-effects in tests.
 
 9. Always use structured logging.  Log any relevant state/context with the
-   messages (but do not log secrets).  If stdout is not a terminal, output
-   the structured logs in jsonl format.
+   messages (but do not log secrets).  If the log stream is not a terminal,
+   output the structured logs in jsonl format.
 
 10. Avoid using bare strings or numbers in code, especially if they appear
     anywhere more than once.  Always define a constant (usually at the top
@@ -102,9 +102,29 @@ Version: 2025-06-08
     build files are acceptable in the root, but source code and other files
     should be organized in appropriate subdirectories.
 
-13. Pre-1.0: NEVER write database migrations. There are no live databases
-    anywhere — every user's local index can be rebuilt from a fresh full
-    backup. When the schema changes, just change `schema.sql` (and any code
-    that touches the affected tables). The local index is disposable until
-    1.0 ships and is tagged.
+13. Pre-1.0: NEVER add a database migration. Migrations are supported, but
+    nothing is installed anywhere yet, so there is nothing to migrate. To
+    change the schema, edit `internal/database/schema/001.sql` (and any
+    code that touches the affected tables) directly. After 1.0, each schema
+    change is a new numbered file in that directory and a released file is
+    never edited; an existing local database is then migrated when vaultik
+    is updated. Before 1.0, a local database left on an older schema is
+    deleted and re-created by a full backup. See
+    [`docs/DATAMODEL.md`](docs/DATAMODEL.md#schema-migrations).
 
+14. Never use `git add -A`.  Stage only the files you intentionally
+    changed.
+
+15. Commit messages carry no attribution or advertising trailers for the
+    tool that helped write the code, or for its vendor: the owner is the
+    sole author of code written with a tool.
+
+16. Run the whole test suite with `make test` every time, and read its full
+    output.  Never run `go test`, a single test or a single package, and
+    never grep the output.
+
+17. Do not stop working on a task until the definition of done given in the
+    initial instruction is met: all of the work, not part or most of it.
+
+18. For estimates: backing up over 2.5Gbit/s ethernet to an S3 server
+    backed by 2000MB/sec SSD takes about 4 seconds per gigabyte.

ARCHITECTURE.md

@@ -1,379 +1,401 @@
-# Vaultik Architecture
+# vaultik architecture
 
-This document describes the internal architecture of Vaultik, focusing on the data model, type instantiation, and the relationships between core modules.
+This document describes how the `vaultik` code is organised. It covers three
+things:
 
-## Overview
+- the packages
+- how they are wired together
+- what happens during a backup, a restore, a verify and a prune
 
-Vaultik is a backup system that uses content-defined chunking for deduplication and packs chunks into large, compressed, encrypted blobs for efficient cloud storage. The system is built around dependency injection using [uber-go/fx](https://github.com/uber-go/fx).
+Two other documents cover the stored data:
 
-## Data Flow
+- [docs/DATAMODEL.md](docs/DATAMODEL.md) describes the local index.
+- [docs/REPOSTRUCTURE.md](docs/REPOSTRUCTURE.md) describes the layout of the
+  destination.
+
+## Terms
+
+- **Snapshot**: one backup of the paths listed under one name in the config.
+- **Chunk**: a piece of a file's content. A chunk is named by the SHA-256 hash
+  of its content.
+- **Blob**: one object at the destination. It holds many chunks, compressed and
+  then encrypted.
+- **Local index**: the SQLite database on the backed-up machine.
+- **Destination**: the place the backups are stored.
+- **Manifest**: the unencrypted list of the blobs one snapshot uses.
+- **Remote key**: a hash of the snapshot ID. It names the snapshot's directory
+  at the destination.
+
+## How a backup moves data
 
 ```
-Source Files
+source files
      │
      ▼
-┌─────────────────┐
-│    Scanner      │  Walks directories, detects changed files
-└────────┬────────┘
+  Scanner    walks the paths and finds the files that changed
      │
      ▼
-┌─────────────────┐
-│    Chunker      │  Splits files into variable-size chunks (FastCDC)
-└────────┬────────┘
+  Chunker    cuts each changed file into chunks
      │
      ▼
-┌─────────────────┐
-│    Packer       │  Accumulates chunks, compresses (zstd), encrypts (age)
-└────────┬────────┘
+  Packer     packs new chunks into a blob, compressed and encrypted
      │
      ▼
-┌─────────────────┐
-│   S3 Client     │  Uploads blobs to remote storage
-└─────────────────┘
+  Storer     uploads the blob to the destination
 ```
 
-## Data Model
+## Packages
 
-### Core Entities
+- `cmd/vaultik`: the `main` function. It starts optional profiling and calls
+  `cli.Entry`.
+- `internal/cli`: the commands, their flags and their help text. It builds the
+  application, takes the lock and handles interrupts.
+- `internal/vaultik`: the code behind each command. `Vaultik` holds every
+  dependency a command uses.
+- `internal/snapshot`: the `Scanner`, which backs up one path, and the
+  `SnapshotManager`, which creates snapshot records and uploads their metadata.
+- `internal/chunker`: cuts a stream into chunks with the FastCDC algorithm.
+- `internal/blob`: the `Packer`, which collects chunks into blobs.
+- `internal/blobgen`: the writer that hashes, compresses and encrypts a blob,
+  and the reader that reverses it.
+- `internal/database`: the local index. It holds the schema and the repository
+  types that read and write the tables.
+  [docs/DATAMODEL.md](docs/DATAMODEL.md#code) says which type serves which
+  table.
+- `internal/storage`: the `Storer` interface and its three implementations, for
+  S3, a local directory and rclone.
+- `internal/s3`: the S3 client that the S3 `Storer` uses.
+- `internal/config`: loads and checks the config file.
+- `internal/ui`: formats the status lines, warnings and errors a user sees.
+- `internal/log`: the logger. It writes to stderr.
+- `internal/pidlock`: the lock that keeps two data-changing commands from
+  running at once.
+- `internal/globals`: the version, commit and commit date, set when the binary
+  is built.
+- `internal/types`: distinct types for IDs, hashes and paths.
 
-The database tracks five primary entities and their relationships:
+## Wiring
 
-```
-┌──────────────┐     ┌──────────────┐     ┌──────────────┐
-│   Snapshot   │────▶│     File     │────▶│    Chunk     │
-└──────────────┘     └──────────────┘     └──────────────┘
-       │                                         │
-       │                                         │
-       ▼                                         ▼
-┌──────────────┐                          ┌──────────────┐
-│     Blob     │◀─────────────────────────│  BlobChunk   │
-└──────────────┘                          └──────────────┘
-```
-
-### Entity Descriptions
-
-#### File (`database.File`)
-Represents a file, directory, or symlink in the backup system. Stores metadata needed for restoration:
-- Path, source_path (for restore path stripping), mtime
-- Size, mode, ownership (uid, gid)
-- Symlink target (if applicable)
-
-#### Chunk (`database.Chunk`)
-A content-addressed unit of data. Files are split into variable-size chunks using the FastCDC algorithm:
-- `ChunkHash`: SHA256 hash of chunk content (primary key)
-- `Size`: Chunk size in bytes
-
-Chunk sizes vary between `avgChunkSize/4` and `avgChunkSize*4` (typically 16KB-256KB for 64KB average).
-
-#### FileChunk (`database.FileChunk`)
-Maps files to their constituent chunks:
-- `FileID`: Reference to the file
-- `Idx`: Position of this chunk within the file (0-indexed)
-- `ChunkHash`: Reference to the chunk
-
-#### Blob (`database.Blob`)
-The final storage unit uploaded to S3. Contains many compressed and encrypted chunks:
-- `ID`: UUID assigned at creation
-- `Hash`: SHA256 of final compressed+encrypted content
-- `UncompressedSize`: Total raw chunk data before compression
-- `CompressedSize`: Size after zstd compression and age encryption
-- `CreatedTS`, `FinishedTS`, `UploadedTS`: Lifecycle timestamps
-
-Blob creation process:
-1. Chunks are accumulated (up to MaxBlobSize, typically 10GB)
-2. Compressed with zstd
-3. Encrypted with age (recipients configured in config)
-4. SHA256 hash computed → becomes filename in S3
-5. Uploaded to `blobs/{hash[0:2]}/{hash[2:4]}/{hash}`
-
-#### BlobChunk (`database.BlobChunk`)
-Maps chunks to their position within blobs:
-- `BlobID`: Reference to the blob
-- `ChunkHash`: Reference to the chunk
-- `Offset`: Byte offset within the uncompressed blob
-- `Length`: Chunk size
-
-#### Snapshot (`database.Snapshot`)
-Represents a point-in-time backup:
-- `ID`: Format is `{hostname}_{snapshot-name}_{RFC3339}` (e.g. `server1_home_2025-06-01T12:00:00Z`)
-- Tracks file count, chunk count, blob count, sizes, compression ratio
-- `CompletedAt`: Null until snapshot finishes successfully
-
-#### SnapshotFile / SnapshotBlob
-Join tables linking snapshots to their files and blobs.
-
-### Relationship Summary
-
-```
-Snapshot 1──────────▶ N SnapshotFile N ◀────────── 1 File
-Snapshot 1──────────▶ N SnapshotBlob N ◀────────── 1 Blob
-File     1──────────▶ N FileChunk    N ◀────────── 1 Chunk
-Blob     1──────────▶ N BlobChunk    N ◀────────── 1 Chunk
-```
-
-## Type Instantiation
-
-### Application Startup
-
-The CLI uses fx for dependency injection. Here's the instantiation order:
+The application is assembled with [uber-go/fx](https://github.com/uber-go/fx), a
+dependency injection library. `cli.NewApp` lists the providers:
 
 ```go
-// cli/app.go: NewApp()
-fx.New(
-    fx.Supply(config.ConfigPath(opts.ConfigPath)),  // 1. Config path
-    fx.Supply(opts.LogOptions),                      // 2. Log options
-    fx.Provide(globals.New),                         // 3. Globals
-    fx.Provide(log.New),                             // 4. Logger config
-    config.Module,                                   // 5. Config
-    database.Module,                                 // 6. Database + Repositories
-    log.Module,                                      // 7. Logger initialization
-    storage.Module,                                  // 8. Storage backend (S3/file/rclone)
-    snapshot.Module,                                 // 9. SnapshotManager + ScannerFactory
-    fx.Provide(vaultik.New),                         // 10. Vaultik orchestrator
-)
+fx.Supply(config.Path(opts.ConfigPath)),
+fx.Supply(opts.LogOptions),
+fx.Provide(globals.New),
+fx.Provide(log.New),
+config.Module,
+database.Module,
+log.Module,
+storage.Module,
+snapshot.Module,
+fx.Provide(vaultik.New),
+fx.Invoke(setupGlobals),
+fx.NopLogger,
 ```
 
-### Key Type Instantiation Points
+Each module provides these:
 
-#### 1. Config (`config.Config`)
-- **Created by**: `config.Module` via `config.LoadConfig()`
-- **When**: Application startup (fx DI)
-- **Contains**: All configuration from YAML file (S3 credentials, encryption keys, paths, etc.)
+- `config.Module` provides `*config.Config`, from `config.Load`.
+- `database.Module` provides `*database.DB` and `*database.Repositories`. It
+  creates the index file with mode `0600` when it does not exist.
+- `storage.Module` provides the `storage.Storer` that the config selects.
+- `snapshot.Module` provides the `snapshot.ScannerFactory` and the
+  `*snapshot.SnapshotManager`.
+- `vaultik.New` provides `*vaultik.Vaultik`.
 
-#### 2. Database (`database.DB`)
-- **Created by**: `database.Module` via `database.New()`
-- **When**: Application startup (fx DI)
-- **Contains**: SQLite connection, path reference
+A command runs through `cli.RunOperation`. It works like this:
 
-#### 3. Repositories (`database.Repositories`)
-- **Created by**: `database.Module` via `database.NewRepositories()`
-- **When**: Application startup (fx DI)
-- **Contains**: All repository interfaces (Files, Chunks, Blobs, Snapshots, etc.)
+- `RunWithApp` takes the lock when the command changes data, and then starts the
+  application.
+- The command's function runs in its own goroutine.
+- When the function returns, it asks the application to shut down.
+- On `SIGINT` or `SIGTERM`, the stop hook cancels the `Vaultik` context. It then
+  waits for the function to return, so that its temporary files are removed.
+- The wait ends after 30 seconds at most.
 
-#### 4. Vaultik (`vaultik.Vaultik`)
-- **Created by**: `vaultik.New(VaultikParams)`
-- **When**: Application startup (fx DI)
-- **Contains**: All dependencies for backup operations
+The `config` commands, `database delete`, `completion` and `version` do not
+build the application.
 
-```go
-type Vaultik struct {
-    Globals         *globals.Globals
-    Config          *config.Config
-    DB              *database.DB
-    Repositories    *database.Repositories
-    Storage         storage.Storer
-    ScannerFactory  snapshot.ScannerFactory
-    SnapshotManager *snapshot.SnapshotManager
-    Shutdowner      fx.Shutdowner
-    Fs              afero.Fs
-    ctx             context.Context
-    cancel          context.CancelFunc
-}
-```
+## Backup
 
-#### 5. SnapshotManager (`snapshot.SnapshotManager`)
-- **Created by**: `snapshot.Module` via `snapshot.NewSnapshotManager()`
-- **When**: Application startup (fx DI)
-- **Responsibility**: Creates/completes snapshots, exports metadata to S3
+`Vaultik.CreateSnapshot` runs a backup. It does these steps in order:
 
-#### 6. Scanner (`snapshot.Scanner`)
-- **Created by**: `ScannerFactory(ScannerParams)`
-- **When**: Each `CreateSnapshot()` call
-- **Contains**: Chunker, Packer, progress reporter
+- `EnsureStorageBinding` checks that the local index belongs to the configured
+  destination.
+- `markDestination` checks the marker file and then writes it.
+- `PruneDatabase` removes unfinished snapshots and unused records from the
+  local index.
+- `createNamedSnapshot` runs once for each snapshot name.
+- With `--prune`, `runPostBackupPrune` removes old snapshots and unused data.
 
-```go
-// vaultik/snapshot.go: CreateSnapshot()
-scanner := v.ScannerFactory(snapshot.ScannerParams{
-    EnableProgress: !opts.Cron,
-    Fs:             v.Fs,
-})
-```
+`CreateSnapshot` calls `skipPostBackupPrune` instead of `runPostBackupPrune` in
+these cases:
 
-#### 7. Chunker (`chunker.Chunker`)
-- **Created by**: `chunker.NewChunker(avgChunkSize)`
-- **When**: Inside `snapshot.NewScanner()`
-- **Configuration**:
-  - `avgChunkSize`: From config (typically 64KB)
-  - `minChunkSize`: avgChunkSize / 4
-  - `maxChunkSize`: avgChunkSize * 4
+- `--skip-errors` skipped a file or directory that could not be read.
+- A configured path held nothing.
+- An earlier snapshot of the same name found a mount point that this backup did
+  not find.
 
-#### 8. Packer (`blob.Packer`)
-- **Created by**: `blob.NewPacker(PackerConfig)`
-- **When**: Inside `snapshot.NewScanner()`
-- **Configuration**:
-  - `MaxBlobSize`: Maximum blob size before finalization (typically 10GB)
-  - `CompressionLevel`: zstd level (1-19)
-  - `Recipients`: age public keys for encryption
+`PruneDatabase` must succeed before anything is scanned. An unfinished snapshot
+left in the index would make the backup skip files that were never fully
+stored.
 
-```go
-// snapshot/scanner.go: NewScanner()
-packerCfg := blob.PackerConfig{
-    MaxBlobSize:      cfg.MaxBlobSize,
-    CompressionLevel: cfg.CompressionLevel,
-    Recipients:       cfg.AgeRecipients,
-    Repositories:     cfg.Repositories,
-    Fs:               cfg.FS,
-}
-packer, err := blob.NewPacker(packerCfg)
-```
+### One snapshot
 
-## Module Responsibilities
+`createNamedSnapshot` does these steps:
 
-### `internal/cli`
-Entry point for fx application. Combines all modules and handles signal interrupts.
+- `resolveSnapshotPaths` makes each configured path absolute and resolves
+  symlinks.
+- `SnapshotManager.CreateSnapshotWithName` creates the snapshot record. The
+  record is not yet marked complete.
+- `Scanner.Scan` backs up each path in turn.
+- `finalizeSnapshotMetadata` uploads the snapshot's metadata and marks the
+  snapshot complete.
 
-Key functions:
-- `NewApp(AppOptions)` → Creates fx.App with all modules
-- `RunApp(ctx, app)` → Starts app, handles graceful shutdown
-- `RunWithApp(ctx, opts)` → Convenience wrapper
+A new `Scanner` is created for each snapshot, from `ScannerFactory`. It owns a
+`Chunker` and a `Packer`.
 
-### `internal/vaultik`
-Main orchestrator containing all dependencies and command implementations.
+### Scanning one path
 
-Key methods:
-- `New(VaultikParams)` → Constructor (fx DI)
-- `CreateSnapshot(opts)` → Main backup operation
-- `ListSnapshots(jsonOutput)` → List available snapshots
-- `VerifySnapshot(id, deep)` → Verify snapshot integrity
-- `PurgeSnapshots(...)` → Remove old snapshots
+`Scanner.Scan` works in these steps:
 
-### `internal/chunker`
-Content-defined chunking using FastCDC algorithm.
+- `repairInterruptedBlobs` deletes the records of blobs whose upload never
+  finished. Their data is then stored again.
+- `loadDatabaseState` loads the known files and chunks into memory.
+- `scanPhase` walks the path. It sorts each entry into changed or unchanged.
+- `detectDeletedFiles` counts the files that the previous snapshot of this name
+  had and that are now gone.
+- `batchAddFilesToSnapshot` records the unchanged files in the new snapshot.
+- `recordDirectories` records every directory, and the mount points found.
+- `processPhase` reads each changed file and stores it.
 
-Key types:
-- `Chunk` → Hash, Data, Offset, Size
-- `Chunker` → avgChunkSize, minChunkSize, maxChunkSize
+The walk handles each kind of entry differently:
 
-Key methods:
-- `NewChunker(avgChunkSize)` → Constructor
-- `ChunkReaderStreaming(reader, callback)` → Stream chunks with callback (preferred)
-- `ChunkReader(reader)` → Return all chunks at once (memory-intensive)
+- A regular file or a symlink is compared with its record in the index.
+- A directory is recorded with its metadata. It has no content to store.
+- A device node, a named pipe or a socket is skipped.
+- An entry that matches an exclude pattern is skipped. For a directory, so is
+  everything in it.
 
-### `internal/blob`
-Blob packing: accumulates chunks, compresses, encrypts, tracks metadata.
+### Deduplication
 
-Key types:
-- `Packer` → Thread-safe blob accumulator
-- `ChunkRef` → Hash + Data for adding to packer
-- `FinishedBlob` → Completed blob ready for upload
-- `BlobWithReader` → FinishedBlob + io.Reader for streaming upload
+Data is stored once at two levels.
 
-Key methods:
-- `NewPacker(PackerConfig)` → Constructor
-- `AddChunk(ChunkRef)` → Add chunk to current blob
-- `FinalizeBlob()` → Compress, encrypt, hash current blob
-- `Flush()` → Finalize any in-progress blob
-- `SetBlobHandler(func)` → Set callback for upload
+- **Files.** A file is not read again when its record in the index still matches
+  it.
+- **Chunks.** A chunk is named by the SHA-256 hash of its content. A chunk whose
+  hash is already in the index is not stored again.
 
-### `internal/snapshot`
+A file's record matches when all of these are unchanged:
 
-#### Scanner
-Orchestrates the backup process for a directory.
+- the size
+- the modification time
+- the mode
+- the owner and group
+- the link target
 
-Key methods:
-- `NewScanner(ScannerConfig)` → Constructor (creates Chunker + Packer)
-- `Scan(ctx, path, snapshotID)` → Main scan operation
+One case overrides the file check. A file is read again when it lists a chunk
+that no uploaded blob holds. This happens when the only snapshot that used the
+blob holding that chunk is removed. `loadFilesToRechunk` finds these files.
 
-Scan phases:
-1. **Phase 0**: Detect deleted files from previous snapshots
-2. **Phase 1**: Walk directory, identify files needing processing
-3. **Phase 2**: Process files (chunk → pack → upload)
+### Chunks
 
-#### SnapshotManager
-Manages snapshot lifecycle and metadata export.
+`chunker.Chunker` cuts a file with the FastCDC algorithm. The cut points depend
+on the content, so an insertion early in a file moves only nearby chunks.
 
-Key methods:
-- `CreateSnapshot(ctx, hostname, version, commit)` → Create snapshot record
-- `CompleteSnapshot(ctx, snapshotID)` → Mark snapshot complete
-- `ExportSnapshotMetadata(ctx, dbPath, snapshotID)` → Export to S3
-- `CleanupIncompleteSnapshots(ctx, hostname)` → Remove failed snapshots
+`chunk_size` sets the average chunk size. No chunk is larger than four times
+this size. Only a file's last chunk can be smaller than a quarter of it. With
+the default of 10 MB, those limits are 40 MB and 2.5 MB.
 
-### `internal/database`
-SQLite database for local index. Single-writer mode for thread safety.
+`ChunkReaderStreaming` passes each chunk to a callback as it is found. The whole
+file is never held in memory.
 
-Key types:
-- `DB` → Database connection wrapper
-- `Repositories` → Collection of all repository interfaces
+### Blobs
 
-Repository interfaces:
-- `FilesRepository` → CRUD for File records
-- `ChunksRepository` → CRUD for Chunk records
-- `BlobsRepository` → CRUD for Blob records
-- `SnapshotsRepository` → CRUD for Snapshot records
-- Plus join table repositories (FileChunks, BlobChunks, etc.)
+`blob.Packer` collects new chunks into a blob. For each blob it does this:
 
-## Snapshot Creation Flow
+- It gives the blob a UUID and creates a temporary file.
+- It adds each chunk's bytes to a running SHA-256 hash.
+- In the same pass it compresses the bytes with zstd and encrypts them with
+  age. The result goes to the temporary file.
+- It records each chunk's offset and length within the blob.
+- It closes the blob when the next chunk would pass `blob_size_limit`, and at
+  the end of each path.
 
-```
-CreateSnapshot(opts)
-    │
-    ├─► CleanupIncompleteSnapshots()   // Critical: avoid dedup errors
-    │
-    ├─► SnapshotManager.CreateSnapshot()   // Create DB record
-    │
-    ├─► For each source directory:
-    │       │
-    │       ├─► scanner.Scan(ctx, path, snapshotID)
-    │       │       │
-    │       │       ├─► Phase 0: detectDeletedFiles()
-    │       │       │
-    │       │       ├─► Phase 1: scanPhase()
-    │       │       │       Walk directory
-    │       │       │       Check file metadata changes
-    │       │       │       Build list of files to process
-    │       │       │
-    │       │       └─► Phase 2: processPhase()
-    │       │               For each file:
-    │       │                   chunker.ChunkReaderStreaming()
-    │       │                   For each chunk:
-    │       │                       packer.AddChunk()
-    │       │                       If blob full → FinalizeBlob()
-    │       │                           → handleBlobReady()
-    │       │                           → s3Client.PutObjectWithProgress()
-    │       │               packer.Flush()  // Final blob
-    │       │
-    │       └─► Accumulate statistics
-    │
-    ├─► SnapshotManager.UpdateSnapshotStatsExtended()
-    │
-    ├─► SnapshotManager.CompleteSnapshot()
-    │
-    └─► SnapshotManager.ExportSnapshotMetadata()
-            │
-            ├─► Copy database to temp file
-            ├─► Clean to only current snapshot data (VACUUM)
-            ├─► Compress binary SQLite with zstd
-            ├─► Encrypt with age
-            ├─► Upload db.zst.age to storage
-            └─► Upload manifest.json.zst to storage
-```
+The blob's name is `hex(SHA256(SHA256(uncompressed contents)))`. It is computed
+from the chunk data before compression and encryption.
+[docs/REPOSTRUCTURE.md](docs/REPOSTRUCTURE.md#blobs) says what follows from
+that.
 
-## Deduplication Strategy
+When a blob is closed, `Scanner.handleBlobReady` stores it:
 
-1. **File-level**: Files unchanged since last backup are skipped (metadata comparison: size, mtime, mode, uid, gid)
+- It looks the blob up at the destination. A blob that is already there is not
+  uploaded again.
+- Otherwise it uploads the temporary file to
+  `blobs/<hash[0:2]>/<hash[2:4]>/<hash>`.
+- It records the upload in the index and links the blob to the snapshot.
+- It deletes the temporary file.
 
-2. **Chunk-level**: Chunks are content-addressed by SHA256 hash. If a chunk hash already exists in the database, the chunk data is not re-uploaded.
+A file's record is written to the index only after every blob that holds its
+chunks is uploaded.
 
-3. **Blob-level**: Blobs contain only unique chunks. Duplicate chunks within a blob are skipped.
+### Snapshot metadata
 
-## Storage Layout in S3
+`finalizeSnapshotMetadata` does these steps in order:
 
-```
-bucket/
-├── blobs/
-│   └── {hash[0:2]}/
-│       └── {hash[2:4]}/
-│           └── {full-hash}          # Compressed+encrypted blob
-│
-└── metadata/
-    └── {snapshot-id}/
-        ├── db.zst.age               # Encrypted binary SQLite database
-        └── manifest.json.zst        # Blob list (for pruning/verification)
-```
+- `PopulateSnapshotBlobs` records every blob the snapshot uses.
+- `UpdateSnapshotStatsExtended` stores the snapshot's totals.
+- `ExportSnapshotMetadata` uploads the metadata.
+- `MarkSnapshotComplete` marks the snapshot complete.
 
-## Thread Safety
+Completion is recorded last on purpose. A backup that stops during the upload
+leaves the snapshot unfinished, and the next backup removes it and stores its
+data again.
 
-- `Packer`: Thread-safe via mutex. Multiple goroutines can call `AddChunk()`.
-- `Scanner`: Uses `packerMu` mutex to coordinate blob finalization.
-- `Database`: Single-writer mode (`MaxOpenConns=1`) ensures SQLite thread safety.
-- `Repositories.WithTx()`: Handles transaction lifecycle automatically.
+`ExportSnapshotMetadata` works on a copy of the index, never on the index
+itself:
+
+- It copies the index into a temporary directory with `VACUUM INTO`.
+- It deletes the rows of every other snapshot from the copy. It also deletes the
+  files and blobs that this snapshot does not use. It then deletes each chunk
+  that no remaining file and no remaining blob uses.
+- It keeps the `local_meta` and `schema_migrations` tables as they are. The
+  uploaded database therefore names the destination, and lists the migrations
+  that have run.
+- It runs `VACUUM`, so that no deleted row is left in the file.
+- It compresses the copy with zstd and encrypts it with age.
+- It builds the manifest from the copy.
+- It uploads `metadata/<remote-key>/db.zst.age`, and then
+  `metadata/<remote-key>/manifest.json.zst`.
+
+The manifest is uploaded last. A directory under `metadata/` with no manifest
+therefore marks a backup that did not finish.
+
+## Restore
+
+`Vaultik.Restore` does these steps:
+
+- `restoreIdentities` parses the age private key. A missing or malformed key
+  stops the restore before anything is downloaded.
+- `downloadSnapshotDB` downloads and decrypts the snapshot's database into a
+  temporary directory.
+- `getFilesToRestore` selects the files, from the paths given.
+- `restoreAllFiles` downloads blobs and writes the files.
+- With `--verify`, `verifyRestoredFiles` reads each restored file and checks its
+  chunk hashes.
+
+The restore does not read the local index or the manifest. It finds each file's
+chunks and blobs through the downloaded database.
+
+### Checks on the downloaded database
+
+Decryption shows that a database is readable. It does not show that the
+database is honest. The restore therefore checks it:
+
+- `database.OpenReadOnly` opens the file read-only. It refuses a file that has a
+  trigger, a view or a virtual table, or that lacks an expected table.
+- `verifySnapshotDBIdentity` checks that the snapshot in the database hashes to
+  the remote key that was asked for.
+- `containedRestorePath` refuses a stored path that would leave the target
+  directory. It also refuses to write through a symlink below the target.
+- A blob is not decompressed beyond the size that the database gives for it.
+- Each downloaded blob is hashed. The hash must match the blob's name.
+
+### Order of work
+
+`restorePlan` orders the files so that few blobs are on disk at once:
+
+- Downloaded blobs are kept decrypted in a cache directory.
+- A file is restored once every blob it needs is in the cache.
+- When no file is ready, the restore downloads the blobs for the file that
+  needs the fewest.
+- `restoreSweeper` deletes a cached blob once no remaining file needs it.
+
+A regular file is created with mode `0600` and gets its stored mode after its
+content is written. A directory gets its stored mode, owner and modification
+time after the whole restore.
+
+## Verify
+
+`Vaultik.VerifySnapshotWithOptions` checks a snapshot without the private key:
+
+- It downloads the manifest.
+- It checks that the snapshot's encrypted database is present.
+- It checks that each listed blob is present with the listed size.
+
+`Vaultik.RunDeepVerify` needs the private key. It does these steps:
+
+- It downloads the manifest and the database. It checks the database as a
+  restore does.
+- `verifyManifestAgainstDatabase` checks that both list the same blobs with the
+  same sizes.
+- `verifyBlobExistenceFromDB` checks that every blob is present with the
+  recorded size.
+- `verifyBlob` downloads each blob, checks the hash of each chunk in it, and
+  checks the hash that names the blob.
+
+## Prune
+
+`Vaultik.Prune` does three steps:
+
+- `CleanupLocalSnapshots` drops from the local index each snapshot whose
+  manifest is gone from the destination.
+- `PruneDatabase` removes unfinished snapshots and unused records from the
+  local index.
+- `PruneBlobs` deletes unused data at the destination.
+
+`PruneBlobs` reads every manifest at the destination. It then deletes two kinds
+of object, once they are older than `prune_grace_period`:
+
+- a blob that no manifest lists
+- a directory under `metadata/` that has no manifest, and whose snapshot is not
+  in the local index
+
+It deletes nothing when any manifest cannot be read.
+[docs/REPOSTRUCTURE.md](docs/REPOSTRUCTURE.md#pruning) explains why this is
+safe, and where it is not.
+
+## The destination
+
+`storage.Storer` is the only way the code reaches the destination. It has these
+methods:
+
+- `Put` and `PutWithProgress` store an object.
+- `Get` reads an object.
+- `Stat` reports an object's size.
+- `Delete` removes an object.
+- `List` and `ListStream` list the objects under a prefix.
+- `DeletePartialUploads` removes what an interrupted upload left behind.
+- `Info` describes the destination.
+
+`storage.NewStorer` picks the implementation from `storage_url`:
+
+- `s3://` gives the S3 implementation.
+- `file://` gives the local directory implementation.
+- `rclone://` gives the rclone implementation.
+
+With no `storage_url`, it builds the S3 implementation from the `s3` settings.
+
+## The marker file and the destination binding
+
+Two checks keep the local index and the destination in step.
+
+- `checkDestinationMarker` looks for `vaultik-destination.txt`. It fails when
+  the file is reported missing while the local index holds snapshots.
+- `EnsureStorageBinding` records the configured destination in the local index
+  on first use. Later it fails when the config names a different destination.
+
+Every command that changes data calls `EnsureStorageBinding` first.
+
+## Concurrency
+
+- The local index is opened with one connection, so its queries run one at a
+  time.
+- The index runs in SQLite's WAL mode. A second process can therefore read it
+  while a backup writes to it.
+- `Packer` guards its state with a mutex.
+- `Scanner` holds `packerMu` while a chunk is added to the packer and while a
+  blob is closed.
+- `Repositories.WithTx` runs a function in one transaction. It commits when the
+  function succeeds and rolls back when it fails.

CLAUDE.md

@@ -1,44 +0,0 @@
-# Rules
-
-Read the rules in AGENTS.md and follow them.
-
-# Memory
-
-* Claude is an inanimate tool.  The spam that Claude attempts to insert into
-  commit messages (which it erroneously refers to as "attribution") is not
-  attribution, as I am the sole author of code created using Claude.  It is
-  corporate advertising for Anthropic and is therefore completely
-  unacceptable in commit messages.
-
-* NEVER use `git add -A`.  Always add only the files you intentionally
-  changed.
-
-* Tests should always be run before committing code.  No commits should be
-  made that do not pass tests.
-
-* Code should always be formatted before committing.  Do not commit
-  unformatted code.
-
-* Code should always be linted before committing.  Do not commit
-  unlinted code.
-
-* The test suite is fast and local.  When running tests, don't run
-  individual parts of the test suite, always run the whole thing by running
-  "make test".
-
-* Do not stop working on a task until you have reached the definition of
-  done provided to you in the initial instruction.  Don't do part or most of
-  the work, do all of the work until the criteria for done are met.
-
-* We do not need to support migrations; schema upgrades can be handled by
-  deleting the local state file and doing a full backup to re-create it.
-
-* When testing on a 2.5Gbit/s ethernet to an s3 server backed by 2000MB/sec SSD, 
-  estimate about 4 seconds per gigabyte of backup time.
-
-* When running tests, don't run individual tests, or grep the output.  run
-  the entire test suite every time and read the full output.
-
-* When running tests, don't run individual tests, or try to grep the output.
-  never run "go test".  only ever run "make test" to run the full test
-  suite, and examine the full output.

Dockerfile

@@ -1,101 +1,101 @@
-# Lint stage
-#
-# This FROM line is the single source of truth for the linter version:
-# script/lint parses the image reference out of it and runs that exact
-# image, so a local `make lint` and CI use the same linter. Bump the
-# linter here (tag AND digest) and nowhere else.
-#
-# golangci/golangci-lint:v2.12.2-alpine, 2026-08-07
-FROM golangci/golangci-lint:v2.12.2-alpine@sha256:91b27804074a0bacea298707f016911e60cf0cdbc6c7bf5ccacb5f0606d18d60 AS lint
-
-RUN apk add --no-cache make build-base
-
-# The context signal for script/lint's native path. This stage runs
-# `make lint` with no docker daemon available, so it is the one place
-# that must run the golangci-lint on PATH directly. script/lint takes
-# that path only when this is set AND the version matches the pin above;
-# version equality alone would also admit a developer's locally
-# installed copy on a host, bypassing the digest pin (issue #80).
-# Nothing outside this stage sets it.
-ENV VAULTIK_LINT_IN_CONTAINER=1
-
+# Lint phase. The linter is invoked directly rather than through `make
+# lint` or `script/lint`, which are themselves a docker build and would
+# recurse into a daemon that does not exist in a build step.
+# golangci/golangci-lint:v2.14.0, 2026-10-05
+FROM golangci/golangci-lint:v2.14.0@sha256:ad862ba6b3798cbe0fd9fd7408d498fd74fbd2623a92406b2fd3898faf0bf98f AS lint
 WORKDIR /src
-
-# Copy go mod files first for better layer caching
 COPY go.mod go.sum ./
-RUN go mod download
-
-# Copy source code
+# Without the time limit a stalled connection to the module proxy holds
+# this step, and every build on the host sharing it, until the kernel
+# drops the connection, which can take an hour. The limit is well above
+# the three minutes a working download can take on a busy build host.
+RUN for attempt in 1 2 3; do \
+        timeout 300 go mod download && exit 0; \
+        echo "go mod download attempt $attempt failed or ran over 300s" >&2; \
+    done; \
+    echo "module download failed: giving up after 3 attempts" >&2; \
+    exit 1
 COPY . .
+# `golangci-lint run` silently ignores an unknown top-level key in
+# .golangci.yml, such as a misspelt `linters:`; `config verify` fails on it.
+RUN golangci-lint config verify --config .golangci.yml
+RUN golangci-lint run --config .golangci.yml ./...
 
-# Run formatting check and linter.
-#
-# CHECK_EPOCH must stay immediately above these RUNs. These layers are
-# keyed on its value, so they are cache-eligible only for a value
-# already built against this same tree. script/cibuild and script/docker
-# each pass a fresh value on every invocation, which is what makes their
-# green mean the checks really executed.
-#
-# The value is expanded into each check command rather than left to a
-# bare declaration, so the cache miss does not depend on BuildKit's
-# unreferenced-ARG handling staying as it is. It also puts the epoch in
-# the build log, where a reader can see the layer was keyed fresh.
-#
-# The guard is what makes a build that omits --build-arg fail instead of
-# lie. An unset ARG is an empty string, and an empty string is a
-# perfectly stable cache key: without the guard the first such build
-# runs the checks and every one after it on an unchanged tree replays
-# these layers from cache, executes nothing, and still exits 0. Failed
-# steps are never cached, so the guard fails on EVERY invocation rather
-# than once -- a bare `docker build .` is now a loud error, not a quiet
-# green. Do not give CHECK_EPOCH a default value; a default would
-# satisfy the guard with a constant and restore the hole.
-#
-# ARG scope is per-stage, so the builder stage declares its own.
-# Everything above this line (apk, go.mod, `go mod download`) is
-# deliberately outside the busted range and keeps caching.
-ARG CHECK_EPOCH
-RUN [ -n "$CHECK_EPOCH" ] || exit 1
-RUN echo "check epoch: ${CHECK_EPOCH}" && make fmt-check
-RUN echo "check epoch: ${CHECK_EPOCH}" && make lint
+# Test phase. -race needs cgo and so a C compiler, which the Debian Go
+# image ships and the alpine one does not.
+# golang:1.26.1 (Debian trixie), 2026-10-05
+FROM golang:1.26.1@sha256:cd78d88e00afadbedd272f977d375a6247455f3a4b1178f8ae8bbcb201743a8a AS test
+WORKDIR /src
+COPY go.mod go.sum ./
+# Limited and retried as in the lint phase above.
+RUN for attempt in 1 2 3; do \
+        timeout 300 go mod download && exit 0; \
+        echo "go mod download attempt $attempt failed or ran over 300s" >&2; \
+    done; \
+    echo "module download failed: giving up after 3 attempts" >&2; \
+    exit 1
+COPY . .
+RUN go test -timeout 90s -race -cover ./... || \
+    { echo "--- Rerunning with -v for details ---"; \
+      go test -timeout 90s -race -v ./...; exit 1; }
 
-# Build stage
+# Build stage. Nothing is wanted from either phase above; the copies
+# are what make BuildKit build them first, so this stage cannot run
+# unless lint and test passed.
 # golang:1.26.1-alpine, 2026-03-17
 FROM golang:1.26.1-alpine@sha256:2389ebfa5b7f43eeafbd6be0c3700cc46690ef842ad962f6c5bd6be49ed82039 AS builder
-
-# Depend on lint stage passing
 COPY --from=lint /src/go.sum /dev/null
-
-ARG VERSION=dev
-
-# Install build dependencies for CGO (mattn/go-sqlite3) and sqlite3 CLI (tests)
-RUN apk add --no-cache make build-base sqlite
-
+COPY --from=test /src/go.sum /dev/null
+RUN apk add --no-cache git
+# A tar-stream context keeps the sender's file owners, which git refuses.
+RUN git config --system --add safe.directory /src
 WORKDIR /src
-
-# Copy go mod files first for better layer caching
 COPY go.mod go.sum ./
-RUN go mod download
-
-# Copy source code
+# Limited and retried as in the lint phase above.
+RUN for attempt in 1 2 3; do \
+        timeout 300 go mod download && exit 0; \
+        echo "go mod download attempt $attempt failed or ran over 300s" >&2; \
+    done; \
+    echo "module download failed: giving up after 3 attempts" >&2; \
+    exit 1
 COPY . .
 
-# Run tests. See the CHECK_EPOCH comment in the lint stage for the
-# mechanism; ARG scope is per-stage, so this stage needs its own
-# declaration, its own guard, and its own expansion, and they must stay
-# immediately above the check RUN.
-ARG CHECK_EPOCH
-RUN [ -n "$CHECK_EPOCH" ] || exit 1
-RUN echo "check epoch: ${CHECK_EPOCH}" && make test
+# The VERSION build arg when one is given, otherwise
+# `git describe --tags --always` on the .git in the build context. The
+# commit and its date always come from that .git. With .git present, a
+# version that is still empty, dev or unknown, or a commit or date that
+# is unknown, fails the build: git is missing or could not read the
+# checkout, as when .git is a file pointing outside the context. A
+# context without .git, such as a source export, stamps "dev" and an
+# "unknown" commit and date.
+ARG VERSION
+RUN VERSION="${VERSION:-$(git describe --tags --always || echo dev)}"; \
+    commit="$(git rev-parse HEAD || echo unknown)"; \
+    commit_date="$(git show -s --format=%cs HEAD || echo unknown)"; \
+    if [ -e .git ]; then \
+        case "$VERSION" in ""|dev|unknown) \
+            echo "version is '$VERSION' although .git is present" >&2; \
+            exit 1 ;; \
+        esac; \
+        if [ "$commit" = unknown ] || [ "$commit_date" = unknown ]; then \
+            echo "commit is '$commit' and its date '$commit_date'" \
+                "although .git is present" >&2; \
+            exit 1; \
+        fi; \
+    fi; \
+    globals=sneak.berlin/go/vaultik/internal/globals; \
+    CGO_ENABLED=0 go build -trimpath \
+        -ldflags="-s -w -X ${globals}.Version=${VERSION} \
+            -X ${globals}.Commit=${commit} \
+            -X ${globals}.CommitDate=${commit_date}" \
+        -o /vaultik ./cmd/vaultik
 
-# Build (pure Go, no CGO required since we use modernc.org/sqlite)
-RUN CGO_ENABLED=0 go build -ldflags "-X 'sneak.berlin/go/vaultik/internal/globals.Version=${VERSION}' -X 'sneak.berlin/go/vaultik/internal/globals.Commit=$(git rev-parse HEAD 2>/dev/null || echo unknown)' -X 'sneak.berlin/go/vaultik/internal/globals.CommitDate=$(git show -s --format=%cs HEAD 2>/dev/null || echo unknown)'" -o /vaultik ./cmd/vaultik
-
-# Runtime stage
+# Runtime stage, and the last one: a plain `docker build .` builds this
+# stage's chain and nothing else.
 # alpine:3.21, 2026-02-25
 FROM alpine:3.21@sha256:c3f8e73fdb79deaebaa2037150150191b9dcbfba68b4a46d70103204c53f4709
 
-RUN apk add --no-cache ca-certificates sqlite
+RUN apk add --no-cache ca-certificates
 
 # Copy binary from builder
 COPY --from=builder /vaultik /usr/local/bin/vaultik

Makefile

@@ -1,7 +1,23 @@
 .PHONY: all bootstrap setup check test lint lint-fix fmt fmt-check build clean deps test-coverage local install release release-snapshot docker hooks
 
-# Version number
-VERSION := 1.0.0-rc.1
+# Where script/bootstrap installs Go when the host has none.
+export PATH := $(PATH):$(CURDIR)/.tool/go/bin
+
+# Version number, derived from git by script/version (`git describe
+# --tags --always --dirty`). This used to be a hardcoded
+# constant, which meant every local build claimed to be a release that
+# had never been tagged.
+VERSION := $(shell script/version)
+
+# $(shell) discards exit status, so a script/version that is missing,
+# non-executable or broken would otherwise leave VERSION empty and every
+# binary built here would print "vaultik " with no version at all. A
+# build that cannot determine what it is must not produce an artifact.
+ifeq ($(strip $(VERSION)),)
+$(error script/version produced no version string; a build that cannot \
+determine its version will not be made. Check that script/version exists \
+and is executable)
+endif
 
 # Build variables
 GIT_REVISION := $(shell git rev-parse HEAD 2>/dev/null || echo "unknown")
@@ -27,9 +43,10 @@ setup:
 check:
 	@script/check
 
-# Run tests only. This runs the ENTIRE suite -- there is no separate
-# integration target and no build-tagged subset held back. In
-# particular internal/vaultik/integration_test.go, which does full
+# Run tests only, by building the test phase of the Dockerfile. This
+# runs the ENTIRE suite -- there is no separate integration target and
+# no build-tagged subset held back. In particular
+# internal/vaultik/integration_test.go, which does full
 # chunk -> pack -> encrypt -> upload -> restore round-trips, runs here.
 # A `test-integration` target used to exist and was removed: no file in
 # the repo carried a build tag, so `-tags=integration` selected nothing
@@ -53,7 +70,18 @@ lint:
 lint-fix:
 	@script/lint-fix
 
-# Build binary.
+# Build binary. `build` is the name the org convention reaches for and
+# the one a caller checks the exit code of; `vaultik` is the file rule
+# that does the work, so an unchanged tree still short-circuits.
+#
+# This alias is not decorative. `build` was listed in .PHONY with no
+# rule, and a phony target with no prerequisites and no recipe is
+# already satisfied: `make build` printed "Nothing to be done" and
+# exited 0 without producing a binary (issue #110). Every name in
+# .PHONY needs a rule for that reason; TestPhonyTargetsAllHaveRules in
+# cmd/vaultik keeps it that way.
+build: vaultik
+
 vaultik: internal/*/*.go cmd/vaultik/*.go
 	go build -ldflags "$(LDFLAGS)" -o $@ ./cmd/vaultik
 
@@ -63,17 +91,17 @@ clean:
 	go clean
 
 # Install dependencies. The linter is deliberately not installed here:
-# script/lint runs the digest-pinned golangci-lint image declared by the
-# Dockerfile's lint stage, which is the single source of truth for the
-# linter version. A second, separately pinned copy on PATH could drift
-# from it and make a local `make lint` disagree with CI.
+# script/lint lints by building the lint phase of the Dockerfile, whose
+# FROM line is the single source of truth for the linter version. A
+# second, separately pinned copy on PATH could drift from it and make a
+# local `make lint` disagree with CI.
 deps:
 	go mod download
 
-# Run tests with coverage. -count=1 for the same reason script/test
-# uses it: without it an unchanged package is served from Go's test
-# result cache, and a coverage profile assembled from cached results
-# describes a run that did not happen.
+# Run tests with coverage, on the host. -count=1 because without it an
+# unchanged package is served from Go's test result cache, and a
+# coverage profile assembled from cached results describes a run that
+# did not happen.
 test-coverage:
 	go test -v -count=1 -coverprofile=coverage.out ./...
 	go tool cover -html=coverage.out -o coverage.html
@@ -87,11 +115,11 @@ install: vaultik
 
 # Build and publish release artifacts (linux/darwin × amd64/arm64) via goreleaser.
 release:
-	goreleaser release --clean
+	@script/release
 
 # Dry-run a release build without publishing or tagging.
 release-snapshot:
-	goreleaser release --clean --snapshot
+	@script/release-snapshot
 
 # Build Docker image.
 docker:

REPO_POLICIES.md

@@ -1,6 +1,6 @@
 ---
 title: Repository Policies
-last_modified: 2026-07-06
+last_modified: 2026-10-04
 ---
 
 This document covers repository structure, tooling, and workflow standards. Code
@@ -60,17 +60,28 @@ style conventions are in separate documents:
   prerequisite since nvm requires bash. yarn is then pinned via
   `corepack prepare yarn@<version> --activate`. Never install "latest" or "lts";
   always exact versions. `script/cibuild` runs the CI build: it changes to the
-  repo root and runs `docker build .`; the Gitea workflow calls it. Four further
-  scripts are our own extensions to the standard: `script/check` runs
-  `script/test`, `script/lint`, and `script/fmt-check`; `script/precommit` is
-  what the git pre-commit hook runs, and it calls `script/check`;
-  `script/install-precommit` installs the git pre-commit hook (the `make hooks`
-  target shims to it); and `script/projectname` (literally that filename) simply
-  outputs the project's name. Scripts that need the name call
-  `script/projectname` — e.g. `script/docker` assembles its image tag from it —
-  so those scripts stay byte-identical across all repos. Repo-type-specific
-  pre-commit extras (e.g. `go mod tidy` verification in Go repos) belong in
-  `script/precommit`, not in the hook itself. Model scripts are at
+  repo root, runs `script/bootstrap`, runs `script/check`, and builds the image
+  with the version; the Gitea workflow calls it. **`script/cibuild` runs
+  `script/bootstrap` first**, because the workflow checks out the repo and runs
+  nothing else, while `script/fmt-check` runs the formatter on the host: on a
+  pristine checkout with nothing installed the run dies there, after the
+  containerised gates have passed. **The bootstrap alone is not enough**:
+  `script/bootstrap` installs node and yarn under nvm and leaves neither on the
+  `PATH` of the shell that called it, so a bare `yarn` still exits 127. The host
+  entrypoints that need yarn — `script/fmt` and `script/fmt-check` — therefore
+  source nvm for the pinned node version before invoking it, exactly as
+  `script/bootstrap`'s own install step does. A runner carrying nothing but
+  docker and git then gets through `script/check`. Four further scripts are our
+  own extensions to the standard: `script/check` runs `script/test`,
+  `script/lint` and `script/fmt-check`; `script/precommit` is what the git
+  pre-commit hook runs, and it calls `script/check`; `script/install-precommit`
+  installs the git pre-commit hook (the `make hooks` target shims to it); and
+  `script/projectname` (literally that filename) simply outputs the project's
+  name. Scripts that need the name call `script/projectname` — e.g.
+  `script/docker` assembles its image tag from it — so those scripts stay
+  byte-identical across all repos. Repo-type-specific pre-commit extras (e.g.
+  `go mod tidy` verification in Go repos) belong in `script/precommit`, not in
+  the hook itself. Model scripts are at
   `https://git.eeqj.de/sneak/prompts/raw/branch/main/script/<name>`. The README
   must document the provided scripts in an **Entrypoints** section (see the
   README requirements below).
@@ -89,87 +100,198 @@ style conventions are in separate documents:
   contributor should be able to understand the entire development workflow by
   reading the Makefile.
 
-- Every repo should have a `Dockerfile`. All Dockerfiles must run `make check`
-  as a build step so the build fails if the branch is not green. For non-server
-  repos, the Dockerfile should bring up a development environment and run
-  `make check`. For server repos, `make check` should run as an early build
-  stage before the final image is assembled. Dockerfiles install development
-  prerequisites by running `script/bootstrap` rather than duplicating installs
-  inline; COPY `script/` and the dependency manifests (`package.json` +
-  `yarn.lock`, `go.mod` + `go.sum`, etc.) before running it so the bootstrap
-  layer stays cached until dependencies change.
+- Every repo should have a `Dockerfile`, and it carries the repo's gates: a
+  `lint` phase and a `test` phase, with the final stage depending on both so the
+  image cannot be built unless they pass. For non-server repos the final stage
+  brings up a development environment; for server repos it is the runtime image.
+  The gate phases and the build stage start from their pinned base images and
+  install what those images lack either inline, as the canonical Go `Dockerfile`
+  below does for `git`, or by running `script/bootstrap`, as the `prompts`
+  repo's own `Dockerfile` does for its yarn packages. The development
+  environment stage installs development prerequisites by running
+  `script/bootstrap` rather than duplicating its installs inline. A stage that
+  runs `script/bootstrap` COPYs `script/` and the dependency manifests
+  (`package.json` + `yarn.lock`, `go.mod` + `go.sum`, etc.) before running it.
 
-- **Dockerfiles must use a separate lint stage for fail-fast feedback.** Go
-  repos use a multistage build where linting runs in an independent stage based
-  on the `golangci/golangci-lint` image (pinned by hash). This stage runs
-  `make fmt-check` and `make lint` before the full build begins. The build stage
-  then declares an explicit dependency on the lint stage via
-  `COPY --from=lint /src/go.sum /dev/null`, which forces BuildKit to complete
-  linting before proceeding to compilation and tests. This ensures lint failures
-  surface in seconds rather than minutes, without blocking on dependency
-  download or compilation in the build stage.
+- **Linting and testing run in Docker, as phases of the `Dockerfile`.** There is
+  no separate lint file. `script/lint` and `script/test` each build one phase
+  and nothing else:
 
-    The standard pattern for a Go repo Dockerfile is:
+    ```sh
+    docker build --no-cache --target lint -t "$(script/projectname)-lint" .
+    docker build --no-cache --target test -t "$(script/projectname)-test" .
+    ```
+
+    **A stage that is not the last one in the file is built only when the final
+    stage's chain depends on it, or when `--target` names it.** That is why the
+    two gates are always invoked by name here, and why the final stage carries a
+    `COPY --from=` of a harmless file from each of them: without that edge a
+    plain `docker build .` builds the last stage alone and exits 0 having linted
+    and tested nothing.
+
+    **Every `docker build` in `script/` is tagged**, here and in
+    `script/cibuild` and `script/docker`. An untagged build leaves a dangling
+    image behind on every invocation, on every developer host and every CI
+    runner; a tagged one replaces the previous image.
+
+    Inside a phase the tool is invoked directly — `golangci-lint`, `go test`,
+    `eslint`, `prettier` — never through `make lint` or `script/test`, which are
+    themselves a `docker build` and would recurse into a daemon that does not
+    exist in a build step. Formatting is the exception and stays on the host:
+    `script/fmt` writes the working tree, and `script/fmt-check` is its
+    read-only twin.
+
+    **No lint verdict may come from a host invocation of the linter.** On a
+    shared host golangci-lint reads a result cache keyed on file content rather
+    than location, so a second checkout of the same content is served the first
+    one's findings, and a host-global lock in `$TMPDIR` makes concurrent runs
+    exit non-zero with `parallel golangci-lint is running` — a status a caller
+    cannot tell from real findings. Both have produced wrong verdicts in this
+    org, in both directions. A container has its own cache, its own `TMPDIR` and
+    a digest-pinned binary, so neither is reachable.
+
+- **Any build that runs checks is built with `--no-cache`.** Docker invalidates
+  a `COPY` layer only when the copied content changes, so on an unchanged tree
+  the check `RUN` is served from cache, nothing executes, and the build still
+  exits 0. Every `docker build` in `script/` therefore passes `--no-cache`:
+  `script/lint`, `script/test`, `script/cibuild` and `script/docker` are the
+  four, and there is no fifth — `script/check` runs the two gate phases and
+  `script/fmt-check`, and builds no image of its own. A bare `docker build .` is
+  not evidence that anything ran: a sub-second build reporting success is a
+  cache hit, not a result. Never invalidate by pruning — `docker builder prune`
+  and friends destroy a build cache shared with every other build on the host.
+  When a check is added or changed, prove it works by planting a defect it must
+  catch and watching the run fail on it, then revert the defect. A green run
+  alone shows neither that the check ran nor that it covers what it should.
+
+- **The gate phases are separate stages, and the build stage depends on both.**
+  The lint phase is based on the `golangci/golangci-lint` image (pinned by
+  hash), so lint failures surface in seconds rather than after a full compile,
+  and the test phase is based on the Debian Go image. The canonical Go repo
+  `Dockerfile`:
 
     ```dockerfile
-    # Lint stage — fast feedback on formatting and lint issues
+    # Lint phase
     # golangci/golangci-lint:v2.x.x, YYYY-MM-DD
     FROM golangci/golangci-lint@sha256:... AS lint
     WORKDIR /src
     COPY go.mod go.sum ./
     RUN go mod download
     COPY . .
-    RUN make fmt-check
-    RUN make lint
+    RUN golangci-lint run --config .golangci.yml ./...
 
-    # Build stage
-    # golang:1.x-alpine, YYYY-MM-DD
-    FROM golang@sha256:... AS builder
+    # Test phase. -race needs cgo and so a C compiler, which the Debian Go
+    # image ships and the alpine one does not.
+    # golang:1.x, YYYY-MM-DD
+    FROM golang@sha256:... AS test
     WORKDIR /src
-
-    # Force BuildKit to run the lint stage before proceeding
-    COPY --from=lint /src/go.sum /dev/null
-
     COPY go.mod go.sum ./
     RUN go mod download
     COPY . .
-    RUN make test
+    RUN go test -timeout 90s -race -cover ./... || \
+        { echo "--- Rerunning with -v for details ---"; \
+          go test -timeout 90s -race -v ./...; exit 1; }
 
-    ARG VERSION=dev
-    RUN CGO_ENABLED=0 go build -trimpath \
+    # Build stage. Nothing is wanted from either phase above; the copies
+    # are what make BuildKit build them first, so this stage cannot run
+    # unless lint and test passed.
+    # golang:1.x-alpine, YYYY-MM-DD
+    FROM golang@sha256:... AS builder
+    COPY --from=lint /src/go.sum /dev/null
+    COPY --from=test /src/go.sum /dev/null
+    RUN apk add --no-cache git
+    # A tar-stream context keeps the sender's file owners, which git refuses.
+    RUN git config --system --add safe.directory /src
+    WORKDIR /src
+    COPY go.mod go.sum ./
+    RUN go mod download
+    COPY . .
+
+    # The VERSION build arg when one is given, otherwise
+    # `git describe --tags --always` on the .git in the build context. With
+    # .git present, a version that is still empty, dev or unknown fails the
+    # build: git is missing or could not read the checkout.
+    ARG VERSION
+    RUN VERSION="${VERSION:-$(git describe --tags --always)}"; \
+        if [ -e .git ]; then \
+            case "$VERSION" in ""|dev|unknown) \
+                echo "version is '$VERSION' although .git is present" >&2; \
+                exit 1 ;; \
+            esac; \
+        fi; \
+        CGO_ENABLED=0 go build -trimpath \
             -ldflags="-s -w -X main.Version=${VERSION}" \
             -o /app ./cmd/app/
 
-    # Runtime stage
+    # Runtime stage, and the last one
     FROM alpine@sha256:...
     COPY --from=builder /app /usr/local/bin/app
     ENTRYPOINT ["app"]
     ```
 
     Key points:
-    - The lint stage uses the `golangci/golangci-lint` image directly (it
-      includes both Go and the linter), so there is no need to install the
-      linter separately.
-    - `COPY --from=lint /src/go.sum /dev/null` is a no-op file copy that creates
-      a stage dependency. BuildKit runs stages in parallel by default; without
-      this line, the build stage would not wait for lint to finish and a lint
-      failure might not fail the overall build.
+    - The lint phase uses the `golangci/golangci-lint` image directly (it has
+      both Go and the linter), so nothing needs installing.
+    - `COPY --from=<phase> /src/go.sum /dev/null` is a no-op copy whose only
+      purpose is the ordering edge. BuildKit runs stages in parallel by default,
+      and a stage nothing depends on is not built at all, so without these two
+      lines a red gate would not fail the build.
+    - Keep the runtime stage last, and if you add a stage after it, give it the
+      same two copies. A plain `docker build .` builds the last stage's chain
+      and nothing else.
     - If the project uses `//go:embed` directives that reference build artifacts
-      (e.g. a web frontend compiled in a separate stage), the lint stage must
+      (e.g. a web frontend compiled in a separate stage), the lint phase must
       create placeholder files so the embed directives resolve. Example:
       `RUN mkdir -p web/dist && touch web/dist/index.html web/dist/style.css`.
-      The lint stage should not depend on the actual build output — it exists to
-      fail fast.
-    - If the project requires CGO or system libraries for linting (e.g.
-      `vips-dev`), install them in the lint stage with `apk add`.
-    - The build stage runs `make test` after compilation setup. Tests run in the
-      build stage, not the lint stage, because they may require compiled
-      artifacts or heavier dependencies.
+    - If the project requires CGO or system libraries for linting, install them
+      in the lint phase. The `golangci/golangci-lint` image is Debian-based and
+      has no `apk`, so install with `apt-get` under the Debian package name
+      (`libvips-dev`, where alpine says `vips-dev`), and delete the package
+      lists in the same `RUN`, so the layer does not keep them:
+
+        ```dockerfile
+        RUN apt-get update \
+            && apt-get install -y --no-install-recommends libvips-dev \
+            && rm -rf /var/lib/apt/lists/*
+        ```
+
+    - `.dockerignore` lets `.git` into the build context. It keeps out every git
+      `config` at any depth (`**/.git/config`, `**/.git/modules/**/config`): the
+      repository's own, each submodule's under `.git/modules/`, and that of a
+      submodule keeping its own `.git` directory. `git describe` does not need
+      them, and each can hold a credential: a password in a remote URL, or the
+      token the CI checkout step stores there. A submodule whose name has a
+      `config` segment (`config`, `deploy/config`, `config/lib`) loses its whole
+      git directory to `**/.git/modules/**/config`, and Go's version stamping
+      then fails the build: give it a name without that segment
+      (`git submodule add --name`). The stage that compiles has `git` (the
+      Debian Go image has it; an alpine one needs `apk add --no-cache git`) and
+      takes the version from the `VERSION` build argument when one is given,
+      otherwise from `git describe --tags --always`. That gives the tag on a
+      tagged commit; on a later commit, the tag, the number of commits since it
+      and the short commit (`v1.2.3-4-gabc1234`); and the short commit when no
+      tag is reachable. The stage that compiles also marks its working directory
+      safe for git (`git config --system --add safe.directory /src`): a context
+      sent as a tar stream keeps the sender's file owners, and git refuses a
+      checkout owned by another user, so the version would come out empty.
+      `ARG VERSION` has no default, and the build fails if the context carries
+      `.git` and the version still comes out empty, `dev` or `unknown`. A plain
+      `docker build .` with no build arguments must succeed; a Dockerfile that
+      refuses an empty build argument drops that refusal and keeps the argument.
 
 - Every repo should have a Gitea Actions workflow (`.gitea/workflows/`) that
-  runs `script/cibuild` (which runs `docker build .`) on push. Since the
-  Dockerfile already runs `make check`, a successful build implies all checks
-  pass.
+  runs `script/cibuild` on push, and checks out the repo as its only other step.
+  That script bootstraps, runs the gate phases, and then builds the image, so a
+  successful run means every check passed; a bare `docker build .` does not
+  carry the same guarantee, because its gate phases may come from the cache. The
+  image build is uncached and so runs the gate phases a second time. That is the
+  price of the rule above, and it is worth paying: the image that ships is built
+  from a run of its own gates rather than from a cache entry. A separate
+  workflow limited to `main` by a `branches` list under `on: push` cannot be
+  checked by review: to try a change to it, add the feature branch to that list
+  and push, then remove the branch from the list again before merging. Keep any
+  job in it that publishes behind `if: github.ref_name == 'main'`, so the run
+  from the feature branch publishes nothing.
 
 - Use platform-standard formatters: `black` for Python, `prettier` for
   JS/CSS/Markdown/HTML, `go fmt` for Go. Always use default configuration with
@@ -189,14 +311,21 @@ style conventions are in separate documents:
   module under test to verify it compiles/parses. There is no excuse for
   `make test` to be a no-op.
 
-- `make test` must complete in under 20 seconds. Add a 30-second timeout in the
-  Makefile.
+- `make test` must complete in under 60 seconds. That is the hard cap, and a
+  suite that exceeds it fails. Under 20 seconds is the target. A suite between
+  20 and 60 seconds is still green, but the overage must be filed as an
+  improvement bug against that repo. Add a 90-second timeout to the test
+  invocation (`go test -timeout 90s`). The backstop deliberately sits above the
+  hard cap so that it catches a genuinely hung test rather than a merely slow
+  one.
 
-- **`make test` should use the conditional verbose rerun pattern.** Run tests
-  without `-v` (verbose) first. If tests fail, automatically rerun with `-v` to
-  show full output. This keeps CI logs and `docker build` output clean on
-  success (just package/suite summaries) while providing full diagnostic detail
-  on failure (every test case, every assertion). The general shell pattern:
+- **The test command should use the conditional verbose rerun pattern.** Run
+  tests without `-v` (verbose) first. If tests fail, automatically rerun with
+  `-v` to show full output. This keeps CI logs and `docker build` output clean
+  on success (just package/suite summaries) while providing full diagnostic
+  detail on failure (every test case, every assertion). The command lives in the
+  `test` phase of the `Dockerfile`, since `script/test` builds that phase; the
+  Makefile form below is the same pattern for any repo-local invocation:
 
     ```makefile
     test:
@@ -209,11 +338,26 @@ style conventions are in separate documents:
 
     ```makefile
     test:
-    	@go test -timeout 30s -race -cover ./... || \
+    	@go test -count=1 -timeout 90s -race -cover ./... || \
     		{ echo "--- Rerunning with -v for details ---"; \
-    		  go test -timeout 30s -race -v ./...; exit 1; }
+    		  go test -count=1 -timeout 90s -race -v ./...; exit 1; }
     ```
 
+    `-count=1` is required on both invocations: it defeats Go's test _result_
+    cache, so neither run can report a stored pass in place of running the
+    tests. It leaves the build cache alone, so it costs the runtime of the suite
+    and no recompilation.
+
+    That cache is Go's own, separate from Docker's layer cache. Go stores a
+    passing result in its cache directory (`GOCACHE`), and when the same tests
+    run again on unchanged code it prints that result, marked `(cached)`,
+    without running them. That matters on a developer's machine, where this
+    target runs and the directory lasts from one run to the next. The `test`
+    phase of the `Dockerfile` needs no `-count=1`: its base image holds no
+    result for this repo's tests and nothing before its `go test` step runs a
+    test, so there is nothing to replay. `--no-cache` (above) is what makes that
+    step run on an unchanged tree.
+
     Python example:
 
     ```makefile
@@ -239,10 +383,84 @@ style conventions are in separate documents:
   must be in `.gitignore`. No exceptions.
 
 - `.gitignore` should be comprehensive from the start: OS files (`.DS_Store`),
-  editor files (`.swp`, `*~`), language build artifacts, and `node_modules/`.
-  Fetch the standard `.gitignore` from
-  `https://git.eeqj.de/sneak/prompts/raw/branch/main/.gitignore` when setting up
-  a new repo.
+  editor files (`.swp`, `*~`), in-repo agent scratch directories (`.claude/`),
+  language build artifacts, and `node_modules/`. Fetch the standard `.gitignore`
+  from `https://git.eeqj.de/sneak/prompts/raw/branch/main/.gitignore` when
+  setting up a new repo. These patterns are written to `.gitignore`'s own
+  semantics, in which an unanchored pattern already matches at every depth; they
+  are not a `.dockerignore` and must not be transplanted into one unmodified.
+
+- **`.dockerignore` does not use `.gitignore` semantics, and copying patterns
+  across unmodified leaves secrets in the build context.** Docker matches with
+  `moby/patternmatcher`: `filepath.Match` semantics plus a `**` extension, so
+  `*` does not cross `/` and a pattern without a leading `**/` is anchored at
+  the build-context root. A `.dockerignore` listing `.env`, `*.pem` and `*.key`
+  therefore excludes only the copies at the repository root, while `config/.env`
+  and `certs/server.key` still reach the context and can land in an image layer
+  — which is more dangerous than a short file with no secret patterns at all,
+  because it reads as solved and stops anyone looking. Give every
+  depth-independent pattern the `**/` prefix and leave only genuinely
+  root-anchored entries unprefixed: `.claude`, and the repo's own host-built
+  binary, written `/myapp` and never `**/myapp`, which would also match
+  `cmd/myapp/` and delete the package directory from the context. Matching is
+  case-sensitive, and an ALL-CAPS twin per pattern still misses `Server.Key`, so
+  secret names use character ranges — `**/*.[kK][eE][yY]`, `**/*.[pP][eE][mM]`,
+  and likewise for `.envrc` and the extensionless SSH keys. Where such a pattern
+  also catches something the build needs, re-include it with a negation
+  (`!docs/example.env`); deleting the pattern reopens the exposure for every
+  other file it covers. Fetch the standard `.dockerignore` from
+  `https://git.eeqj.de/sneak/prompts/raw/branch/main/.dockerignore` and extend
+  it with the repo's own artifacts.
+
+- **In-repo agent scratch belongs in both files, written to each file's own
+  semantics.** `.claude/` holds one worktree per in-flight agent — an entire
+  additional checkout of the repo — so under `COPY . .` the build context
+  inflates by a multiple of the repo and another session's unreviewed work can
+  be copied into an image layer. In `.gitignore` the entry is `.claude/`,
+  unanchored. In `.dockerignore` it is `.claude`, anchored and with **no** `**/`
+  prefix, because the prefixed form would also delete any nested directory of
+  that name from the build. Anchoring carries a known gap that the canonical
+  `.dockerignore` states in its own comment, since consuming repos receive the
+  file and not the tracker: the directory is created in the agent's working
+  directory, so a repo running agents in subdirectories still ships
+  `services/api/.claude/` and must add its own anchored entry there.
+
+- **A plain `docker build .` of a clone stamps the version that
+  `git describe --tags --always` gives**, derived from the `.git` in the build
+  context as the canonical `Dockerfile` above shows. Without its failure check,
+  a missing `git` or an unreadable checkout would leave `-X main.Version=` empty
+  and the build would still exit 0. `script/docker` and `script/cibuild` pass
+  the version they compute on the host; it takes precedence. They do this
+  byte-identically across repos:
+
+    ```sh
+    # Own line: a failing command substitution inside an argument does not
+    # trip `set -e`, so the inline form degrades to an empty constant.
+    version="$(git describe --tags --always --dirty 2>/dev/null || true)"
+    [ -n "$version" ] || version="unknown"
+    docker build --no-cache \
+        --build-arg VERSION="$version" \
+        -t "$(script/projectname)" .
+    ```
+
+    `--always` makes an untagged repo yield an abbreviated commit hash rather
+    than failing, and the `[ -n "$version" ]` line is the single place the
+    fallback is applied — a live check that fires on a build from an export with
+    no `.git` and on a repository with no commits yet. Do not fold it into the
+    substitution as `|| echo unknown`, which makes the guard unreachable. The
+    Dockerfile's side is `ARG VERSION` in the stage that compiles, declared
+    there because `ARG` is stage-scoped; passing `VERSION` to a repo whose
+    Dockerfile declares no such `ARG` is ignored and costs nothing, which is why
+    the scripts stay byte-identical. One consequence for CI: the standard
+    checkout action clones shallow and fetches no tags, so a repo that embeds a
+    tag-derived version must set `fetch-depth: 0` on its checkout step.
+
+- **Verify `.dockerignore` by enumerating the image, not by reading the
+  patterns.** Plant files at the root _and_ at least two directories deep, build
+  a probe image that does `COPY . .`, and list what actually landed
+  (`docker run --rm --entrypoint find IMAGE /app`). The `transferring context`
+  size is not a substitute: a nested secret is a few bytes, and BuildKit
+  transfers only the delta from the previous build.
 
 - **No build artifacts in version control.** Code-derived data (compiled
   bundles, minified output, generated assets) must never be committed to the
@@ -258,9 +476,56 @@ style conventions are in separate documents:
 - Make all changes on a feature branch. You can do whatever you want on a
   feature branch.
 
-- `.golangci.yml` is standardized and must _NEVER_ be modified by an agent, only
-  manually by the user. Fetch from
-  `https://git.eeqj.de/sneak/prompts/raw/branch/main/.golangci.yml`.
+- `.golangci.yml` is standardized. The vendored copy in a consuming repo must
+  _NEVER_ be modified by an agent: fetch it from
+  `https://git.eeqj.de/sneak/prompts/raw/branch/main/.golangci.yml` and keep it
+  byte-identical, so that no repo can quietly loosen its own linting. Linter
+  configuration changes are made to the canonical copy in the `prompts` repo and
+  reach consuming repos by re-vendoring; an agent may open a PR against
+  canonical, which only the user merges. One list is exempt from byte-identity,
+  because it cannot be written once for every repo: the `deny` list of the
+  `test-support` depguard rule, where a repo names its own test-support packages
+  by full import path. A repo adds entries there and changes nothing else, and a
+  re-vendor carries its entries forward. The canonical golangci-lint version is
+  v2.14.0 (released 2026-09-24), pinned as the digest of the lint phase's base
+  image
+  (`golangci/golangci-lint@sha256:ad862ba6b3798cbe0fd9fd7408d498fd74fbd2623a92406b2fd3898faf0bf98f`,
+  which reports `2.14.0 built with go1.27.0 from 114493f9`). A module's `go`
+  directive must not name a newer Go minor version than the one golangci-lint
+  was built with, or golangci-lint refuses to lint it: this release lints
+  `go 1.27.1` but not `go 1.28`. That digest is the only pin, since no repo
+  installs golangci-lint on the host. A repo sets the lint phase digest to the
+  one named here and re-vendors `.golangci.yml` in the same commit, whichever of
+  the two prompted the change: the canonical copy can name linters that an older
+  golangci-lint rejects, and a newer golangci-lint can add linters that
+  `default: all` switches on until the canonical copy disables them.
+
+- **`script/bootstrap` installs a pinned tool by comparing versions, never by
+  testing presence.** An `if ! command -v <tool>; then install; fi` guard tests
+  `PATH` only, so on an already-provisioned machine the pin is inert and a
+  version bump is a silent no-op — while the Dockerfile, installing into a clean
+  image, gets the pinned version, so a local `make check` and `make docker` can
+  disagree about what the tool even is. The canonical form:
+    - compares the installed version against the pin over the **whole** version
+      token; a parser that stops at the first `-` reports `2.12.2` for a host
+      running `2.12.2-rc1` and skips the install;
+    - treats absent, non-zero, empty or unrecognised `--version` output as a
+      mismatch, so the failure direction is a redundant install and never a
+      skipped one;
+    - after installing, re-resolves the binary the way callers do — `hash -r`,
+      then through `PATH`, not through the directory the installer wrote to —
+      and fails naming the resolved path, since an install that a shadowing
+      binary hides succeeds while changing nothing any caller sees;
+    - is actually called, and prints the version on both success paths: a
+      function defined and never invoked has the same exit status and the same
+      empty output as one that worked.
+
+    Keep it POSIX sh: no arrays, no `[[`, no `grep -P`.
+
+    A Go tool a repo needs on the host is installed with `go install` pinned to
+    a commit hash (`go install <package>@<commit hash>`). It is never tracked as
+    a `go.mod` tool dependency or through a `tools.go` file, either of which
+    pulls the tool's own dependencies into the repo's `go.mod` and `go.sum`.
 
 - When pinning images or packages by hash, add a comment above the reference
   with the version and date (YYYY-MM-DD).
@@ -374,12 +639,14 @@ style conventions are in separate documents:
   settings.
 
 - Avoid putting files in the repo root unless necessary. Root should contain
-  only project-level config files (`README.md`, `Makefile`, `Dockerfile`,
-  `LICENSE`, `.gitignore`, `.editorconfig`, `REPO_POLICIES.md`, and
-  language-specific config). Everything else goes in a subdirectory. Canonical
-  subdirectory names:
+  only project-level config files (`README.md`, `AGENTS.md`, `Makefile`,
+  `Dockerfile`, `LICENSE`, `.gitignore`, `.editorconfig`, `REPO_POLICIES.md`,
+  and language-specific config). Everything else goes in a subdirectory.
+  Canonical subdirectory names:
     - `bin/` — executable scripts and tools
-    - `cmd/` — Go command entrypoints
+    - `cmd/` — Go command entrypoints; thin only: one `main.go` per binary whose
+      body is a single call into `internal/` or `pkg/`, no project logic in
+      `cmd/`
     - `configs/` — configuration templates and examples
     - `deploy/` — deployment manifests (k8s, compose, terraform)
     - `docs/` — documentation and markdown (README.md stays in root)
@@ -406,3 +673,7 @@ style conventions are in separate documents:
     - Go: `go.mod`, `go.sum`, `.golangci.yml`
     - JS: `package.json`, `yarn.lock`, `.prettierrc`, `.prettierignore`
     - Python: `pyproject.toml`
+
+- Guidance for coding agents lives in one `AGENTS.md` at the repository root. It
+  is never committed under a file or directory named after one agent tool, such
+  as `CLAUDE.md` or `.claude/`, and never split into separate memory files.

cmd/vaultik/dockermoddownload_test.go

@@ -0,0 +1,45 @@
+package main_test
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// TestDockerfileBoundsEveryModuleDownload fails unless every
+// `go mod download` in the Dockerfile runs under a time limit, is tried
+// a fixed number of times, and then fails the build with a message that
+// names the module download (issue #351). Like the version tests, this
+// parses the Dockerfile; that a stalled download really is cut off and
+// retried was checked by hand.
+func TestDockerfileBoundsEveryModuleDownload(t *testing.T) {
+	t.Parallel()
+
+	var downloads []string
+
+	for _, instruction := range instructions(t, productDockerfile) {
+		if strings.Contains(instruction, "go mod download") {
+			downloads = append(downloads, instruction)
+		}
+	}
+
+	require.NotEmpty(t, downloads, "%s must download modules",
+		productDockerfile)
+
+	for _, download := range downloads {
+		assert.Contains(t, download, "timeout 300 go mod download",
+			"%s: a module download must have a time limit: %s",
+			productDockerfile, download)
+		assert.Contains(t, download, "for attempt in 1 2 3",
+			"%s: a module download must be retried: %s",
+			productDockerfile, download)
+		assert.Contains(t, download, "module download failed",
+			"%s: a module download that never succeeds must say so: %s",
+			productDockerfile, download)
+		assert.True(t, strings.HasSuffix(download, "exit 1"),
+			"%s: a module download that never succeeds must fail the"+
+				" build: %s", productDockerfile, download)
+	}
+}

cmd/vaultik/dockerversion_test.go

@@ -0,0 +1,149 @@
+package main_test
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// This file guards the version stamping of the product image (issue
+// #75). The failure it protects against is silent: the image still
+// builds and runs, but `vaultik version` inside it reports "commit:
+// unknown" or a version of "dev", so an operator cannot tell which
+// source produced a given backup. The build takes the version as a
+// build arg, which script/docker computes on the host, and otherwise
+// derives it from the .git in its context; the commit and its date
+// always come from that .git.
+//
+// These are parses of the committed files, because shelling out to
+// docker would nest a build inside `make test`. That `vaultik version`
+// in the built image really prints the host's version is verified by
+// hand.
+
+// The files under guard, relative to the repository root.
+const (
+	productDockerfile = "Dockerfile"
+	dockerScript      = "script/docker"
+)
+
+// TestProductDockerfileTakesVersionAsBuildArg fails unless the build
+// declares ARG VERSION, with no default, and stamps it into the binary
+// whenever it is given, ahead of the value derived in the container.
+func TestProductDockerfileTakesVersionAsBuildArg(t *testing.T) {
+	t.Parallel()
+
+	found := instructions(t, productDockerfile)
+
+	require.Contains(t, found, "ARG VERSION",
+		"%s must declare `ARG VERSION`, with no default, so the host can"+
+			" pass it in", productDockerfile)
+
+	assertLdflagReferences(t, found, "VERSION")
+}
+
+// TestProductDockerfileDerivesVersionFromGit fails unless a build given
+// no VERSION, such as a plain `docker build .` of a clone, takes it from
+// `git describe` of the .git in its context, stamps "dev" when the
+// context has no .git, and fails rather than stamp "dev" when that .git
+// yields no version. The commit and its date come from the same .git,
+// and the build fails rather than stamp them "unknown" when it is
+// present.
+func TestProductDockerfileDerivesVersionFromGit(t *testing.T) {
+	t.Parallel()
+
+	found := instructions(t, productDockerfile)
+
+	buildAt := indexContaining(found, "go build")
+	require.GreaterOrEqual(t, buildAt, 0, "%s must build", productDockerfile)
+
+	assert.Contains(t, found[buildAt], "git describe --tags --always || echo dev",
+		"%s must derive the version from git when no VERSION is given,"+
+			" and stamp dev when the context has no .git", productDockerfile)
+	assert.Contains(t, found[buildAt], "[ -e .git ]",
+		"%s must fail when the context carries .git but yields no version",
+		productDockerfile)
+	assert.Contains(t, found[buildAt], "git rev-parse HEAD",
+		"%s must stamp the commit from git", productDockerfile)
+	assert.Contains(t, found[buildAt], "git show -s --format=%cs HEAD",
+		"%s must stamp the commit date from git", productDockerfile)
+	assert.Contains(t, found[buildAt],
+		`[ "$commit" = unknown ] || [ "$commit_date" = unknown ]`,
+		"%s must fail when the context carries .git but yields no commit"+
+			" or date", productDockerfile)
+}
+
+// TestDockerScriptComputesVersionOnTheHost fails unless script/docker
+// passes the version it derives where .git exists as a build arg.
+func TestDockerScriptComputesVersionOnTheHost(t *testing.T) {
+	t.Parallel()
+
+	script := readRepoFile(t, dockerScript)
+
+	assert.Contains(t, script, "--build-arg VERSION=",
+		"%s must pass --build-arg VERSION to the build", dockerScript)
+}
+
+// assertLdflagReferences fails unless the build instruction uses the
+// named ARG whenever it is given (a ${arg:- reference), so a value
+// passed in is not overridden by one derived inside the container.
+func assertLdflagReferences(t *testing.T, found []string, arg string) {
+	t.Helper()
+
+	for _, instruction := range found {
+		if strings.HasPrefix(instruction, "RUN ") &&
+			strings.Contains(instruction, "go build") &&
+			strings.Contains(instruction, "${"+arg+":-") {
+			return
+		}
+	}
+
+	assert.Fail(t, "version arg is declared but never stamped",
+		"the go build in %s must use ${%s:-...}, or the arg is passed and"+
+			" discarded", productDockerfile, arg)
+}
+
+// instructions returns the Dockerfile's instructions, one per element,
+// with comments and blank lines dropped and continuation lines joined,
+// so a multi-line RUN is one string.
+func instructions(t *testing.T, name string) []string {
+	t.Helper()
+
+	var (
+		out         []string
+		continued   string
+		isContinued bool
+	)
+
+	for line := range strings.SplitSeq(readRepoFile(t, name), "\n") {
+		trimmed := strings.TrimSpace(line)
+		if !isContinued && (trimmed == "" || strings.HasPrefix(trimmed, "#")) {
+			continue
+		}
+
+		isContinued = strings.HasSuffix(trimmed, `\`)
+		continued += strings.TrimSuffix(trimmed, `\`)
+
+		if isContinued {
+			continue
+		}
+
+		out = append(out, strings.Join(strings.Fields(continued), " "))
+		continued = ""
+	}
+
+	return out
+}
+
+// indexContaining returns the position of the first instruction
+// containing want; -1 if there is none.
+func indexContaining(found []string, want string) int {
+	for i, instruction := range found {
+		if strings.Contains(instruction, want) {
+			return i
+		}
+	}
+
+	return -1
+}

cmd/vaultik/main.go

@@ -10,6 +10,16 @@ import (
 )
 
 func main() {
+	os.Exit(run())
+}
+
+// run sets up optional profiling, runs the CLI, and returns the process
+// exit code. os.Exit lives in main so it fires only after run's deferred
+// profile writers have flushed. cli.Entry returns a status code rather
+// than calling os.Exit itself: an os.Exit from inside it would skip
+// these defers and truncate the profile of a failing command -- exactly
+// the command one most often wants to profile.
+func run() int {
 	// CPU profiling: set VAULTIK_CPUPROFILE=/path/to/cpu.prof
 	if cpuProfile := os.Getenv("VAULTIK_CPUPROFILE"); cpuProfile != "" {
 		f, err := os.Create(cpuProfile) //nolint:gosec // G304: operator-set path
@@ -46,5 +56,5 @@ func main() {
 		}()
 	}
 
-	cli.Entry()
+	return cli.Entry()
 }

cmd/vaultik/makefile_test.go

@@ -0,0 +1,187 @@
+package main_test
+
+import (
+	"os"
+	"path/filepath"
+	"regexp"
+	"slices"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// This file guards the Makefile that builds this program, which is why
+// it lives beside it rather than in a package of its own.
+//
+// Issue #110: `build` was listed in .PHONY with no `build:` rule
+// anywhere in the file. That combination is silently successful — make
+// considers a phony target with no prerequisites and no recipe already
+// satisfied, so `rm -f vaultik && make build` printed "Nothing to be
+// done for 'build'" and exited 0 with no binary produced. Declaring the
+// name phony is precisely what converts the "No rule to make target"
+// error into a green.
+//
+// The guard is a parse of the Makefile rather than an invocation of
+// make. `make test` is what runs these tests, so shelling back into
+// `make build` here would nest a build inside the test run and drop a
+// binary into the tree as a side effect of testing. The one property a
+// parse cannot establish — that the recipe still fails when the build
+// fails — is not testable from inside the build either; it is verified
+// by hand against a deliberately broken tree.
+
+// phonyDirective introduces the list of phony target names.
+const phonyDirective = ".PHONY:"
+
+// ruleLine matches a rule's target list: a target starts in column
+// zero, so recipe lines (tab-indented) and the continuation lines of a
+// variable assignment (space-indented) are excluded by construction.
+//
+// The trailing (?:[^=]|$) rejects `:=` assignments such as
+// `VERSION := $(shell script/version)`, which are not rules. Directives
+// and function calls (`.PHONY:`, `ifeq`, `$(error ...)`) do not match
+// because a target here must begin with a letter, digit or underscore.
+var ruleLine = regexp.MustCompile(`^([A-Za-z0-9_][A-Za-z0-9_./ -]*):(?:[^=]|$)`)
+
+// TestPhonyTargetsAllHaveRules fails on any name in .PHONY that has no
+// rule in the Makefile. Such a name is not a build target at all: it is
+// a command that reports success without doing anything, which is worse
+// than one that does not exist, because a caller checking the exit code
+// cannot tell the difference.
+func TestPhonyTargetsAllHaveRules(t *testing.T) {
+	t.Parallel()
+
+	makefile := readMakefile(t)
+
+	phony := phonyTargets(makefile)
+	require.NotEmpty(t, phony, "no .PHONY names found; the parser is broken")
+
+	rules := declaredRules(makefile)
+
+	// Sanity check on the rule parser before trusting its verdict: a
+	// parser that found nothing would pass this test by accident.
+	require.Contains(t, rules, "vaultik",
+		"the file rule that builds the binary must be recognized")
+
+	for _, target := range phony {
+		assert.Contains(t, rules, target,
+			"`.PHONY` lists %q but the Makefile declares no %q rule, so "+
+				"`make %s` exits 0 without doing anything", target, target, target)
+	}
+}
+
+// TestBuildTargetBuildsTheBinary pins the specific shape of issue #110:
+// `make build` has to reach the rule that produces the binary. The test
+// above would also pass if `build:` were given an empty recipe of its
+// own, which would be the same silent success under a different
+// spelling.
+func TestBuildTargetBuildsTheBinary(t *testing.T) {
+	t.Parallel()
+
+	prerequisites := rulePrerequisites(readMakefile(t), "build")
+	require.NotNil(t, prerequisites, "the Makefile declares no `build` rule")
+
+	assert.Contains(t, prerequisites, "vaultik",
+		"`make build` must depend on the rule that builds the binary")
+}
+
+// readMakefile returns the contents of the repository's Makefile.
+func readMakefile(t *testing.T) string {
+	t.Helper()
+
+	return readRepoFile(t, "Makefile")
+}
+
+// readRepoFile reads a file by its path relative to the repository
+// root.
+func readRepoFile(t *testing.T, name string) string {
+	t.Helper()
+
+	//nolint:gosec // G304: the path is a constant relative to this repo
+	contents, err := os.ReadFile(filepath.Join(repoRoot(t), name))
+	require.NoError(t, err)
+
+	return string(contents)
+}
+
+// repoRoot returns the repository root. The test binary runs with its
+// package directory as the working directory, so the root is found by
+// walking up until the module file appears.
+func repoRoot(t *testing.T) string {
+	t.Helper()
+
+	dir, err := os.Getwd()
+	require.NoError(t, err)
+
+	for {
+		_, err = os.Stat(filepath.Join(dir, "go.mod"))
+		if err == nil {
+			return dir
+		}
+
+		parent := filepath.Dir(dir)
+		require.NotEqual(t, dir, parent,
+			"walked to the filesystem root without finding a go.mod")
+
+		dir = parent
+	}
+}
+
+// phonyTargets returns every name declared phony, across all .PHONY
+// lines.
+func phonyTargets(makefile string) []string {
+	var targets []string
+
+	for line := range strings.SplitSeq(makefile, "\n") {
+		if !strings.HasPrefix(line, phonyDirective) {
+			continue
+		}
+
+		targets = append(targets,
+			strings.Fields(strings.TrimPrefix(line, phonyDirective))...)
+	}
+
+	return targets
+}
+
+// declaredRules returns the set of target names that have a rule.
+func declaredRules(makefile string) map[string]bool {
+	rules := make(map[string]bool)
+
+	for line := range strings.SplitSeq(makefile, "\n") {
+		match := ruleLine.FindStringSubmatch(line)
+		if match == nil {
+			continue
+		}
+
+		// One rule may name several targets: `a b: prereq`.
+		for target := range strings.FieldsSeq(match[1]) {
+			rules[target] = true
+		}
+	}
+
+	return rules
+}
+
+// rulePrerequisites returns the prerequisites of the named rule, or nil
+// if no such rule exists. A rule with none returns an empty slice, so
+// "declared with nothing to do" is distinguishable from "not declared".
+func rulePrerequisites(makefile, target string) []string {
+	for line := range strings.SplitSeq(makefile, "\n") {
+		match := ruleLine.FindStringSubmatch(line)
+		if match == nil {
+			continue
+		}
+
+		if !slices.Contains(strings.Fields(match[1]), target) {
+			continue
+		}
+
+		_, after, _ := strings.Cut(line, ":")
+
+		return append([]string{}, strings.Fields(after)...)
+	}
+
+	return nil
+}

config.example.yml

@@ -1,15 +1,31 @@
 # vaultik configuration file example
-# This file shows all available configuration options with their default values
-# Copy this file and uncomment/modify the values you need
+# This file shows every configuration option. The values are examples, not
+# defaults; where an option has a default, its comment gives it.
+# Copy this file and change the values you need
 
 # Age recipient public keys for encryption
-# This is REQUIRED - backups are encrypted to these public keys
-# Generate with: age-keygen | grep "public key"
+# Backups are encrypted to these public keys. snapshot create needs at least
+# one; listing, verifying and restoring do not
+# Generate a key pair, and show its public key, with:
+#   age-keygen -o vaultik_backup_private_key.txt
+#   grep 'public key' vaultik_backup_private_key.txt
+# Keep the private key file: restoring needs it. Do not pipe age-keygen
+# straight into grep, which would discard the private key
 age_recipients:
   - age1cj2k2addawy294f6k2gr2mf9gps9r3syplryxca3nvxj3daqm96qfp84tz
 
+# Age private key for decryption
+# Only snapshot restore and snapshot verify --deep need it. The
+# VAULTIK_AGE_SECRET_KEY environment variable overrides this setting
+# Do not set it on a machine you back up: prefer the environment variable,
+# supplied only on the machine you restore on
+# Default: not set
+#age_secret_key: YOUR_AGE_PRIVATE_KEY
+
 # Named snapshots - each snapshot can contain multiple paths
 # Each snapshot gets its own ID and can have snapshot-specific excludes
+# Required: at least one snapshot with at least one path. Only the config
+# commands, version and completion run without one
 snapshots:
   testing:
     paths:
@@ -257,16 +273,21 @@ exclude:
 
 # Storage URL - use either this OR the s3 section below
 # Supports: s3://bucket/prefix, file:///path, rclone://remote/path
-storage_url: "rclone://las1stor1//srv/pool.2024.04/backups/heraklion"
+# An s3:// URL also needs s3.access_key_id and s3.secret_access_key below
+# A file:// path must be absolute, and the directory must already exist:
+# vaultik never creates it
+# An rclone:// URL needs the remote to be defined in rclone's own config
+# (rclone config)
+storage_url: "rclone://myremote/path/to/backups"
 
 # S3-compatible storage configuration
 #s3:
 #  # S3-compatible endpoint URL
 #  # Examples: https://s3.amazonaws.com, https://storage.googleapis.com
-#  endpoint: http://10.100.205.122:8333
+#  endpoint: https://s3.example.com
 #
 #  # Bucket name where backups will be stored
-#  bucket: testbucket
+#  bucket: mybucket
 #
 #  # Prefix (folder) within the bucket for this host's backups
 #  # Useful for organizing backups from multiple hosts
@@ -274,36 +295,54 @@ storage_url: "rclone://las1stor1//srv/pool.2024.04/backups/heraklion"
 #  #prefix: "hosts/myserver/"
 #
 #  # S3 access credentials
-#  access_key_id: Z9GT22M9YFU08WRMC5D4
-#  secret_access_key: Pi0tPKjFbN4rZlRhcA4zBtEkib04yy2WcIzI+AXk
+#  access_key_id: YOUR_ACCESS_KEY
+#  secret_access_key: YOUR_SECRET_KEY
 #
 #  # S3 region
 #  # Default: us-east-1
 #  #region: us-east-1
 #
 #  # Use SSL/TLS for S3 connections
+#  # Only read when storage_url is not set and the endpoint has no scheme;
+#  # with an s3:// storage_url, use ?ssl=false in the URL instead
 #  # Default: true
 #  #use_ssl: true
 #
 #  # Part size for multipart uploads
-#  # Minimum 5MB, affects memory usage during upload
-#  # Supports: 5MB, 10M, 100MiB, etc.
-#  # Default: 5MB
-#  #part_size: 5MB
+#  # Minimum 5MiB, maximum 5GiB; affects memory usage during upload
+#  # A blob too large for 10,000 parts of this size gets larger parts
+#  # Supports: 10MB, 16MiB, 100MiB, etc. (5MB is below the minimum)
+#  # Default: 5MiB
+#  #part_size: 5MiB
 
 # Path to local SQLite index database
 # This database tracks file state for incremental backups
-# Default: /var/lib/vaultik/index.sqlite
-#index_path: /var/lib/vaultik/index.sqlite
+# The VAULTIK_INDEX_PATH environment variable overrides this setting
+# Default: the platform data directory, e.g.
+#   macOS: ~/Library/Application Support/vaultik/index.sqlite
+#   Linux: ~/.local/share/vaultik/index.sqlite
+#index_path: /path/to/index.sqlite
 
 # Average chunk size for content-defined chunking
 # Smaller chunks = better deduplication but more metadata
-# Supports: 10MB, 5M, 1GB, 500KB, 64MiB, etc.
+# Minimum 1MB
+# Supports: 10MB, 5M, 1GB, 64MiB, etc.
 # Default: 10MB
 #chunk_size: 10MB
 
 # Maximum blob size
 # Multiple chunks are packed into blobs up to this size
+# Must be at least four times chunk_size (the largest chunk the chunker can
+# emit); a smaller limit would let a single-chunk blob exceed it.
+# Chunking uses no secret (the FastCDC parameters are fixed and public). At a
+# large limit a blob holds hundreds of chunks, so individual chunk lengths are
+# not visible in its size; lowering the limit toward chunk_size exposes them.
+# A backup needs free temporary space, because each blob is written whole to
+# a temporary file before it is uploaded (up to about blob_size_limit; an
+# rclone destination that cannot stream uploads needs about twice that) and
+# the metadata export writes copies of the local index. Temporary files go to
+# $TMPDIR (default /tmp); with TMPDIR unset, SQLite writes one of those copies
+# to /var/tmp.
 # Supports: 1GB, 10G, 500MB, 1GiB, etc.
 # Default: 10GB
 #blob_size_limit: 10GB
@@ -312,6 +351,20 @@ storage_url: "rclone://las1stor1//srv/pool.2024.04/backups/heraklion"
 # Higher = better compression but slower
 # Default: 3
 compression_level: 5
+
+# How old a blob that no snapshot lists must be before prune deletes it, and
+# how old every object in a directory under metadata/ with no manifest must
+# be before prune deletes that directory
+# A backup uploads its blobs and its db.zst.age before its manifest, so on a
+# destination shared between hosts or users this must be longer than the
+# longest backup run
+# It does not protect a blob a backup reuses from an earlier run that stopped
+# before its manifest: that blob keeps the time of its first upload, so a
+# prune can delete it
+# Supports: 36h, 90m, etc.; 0s deletes such blobs and directories at once
+# Default: 24h
+#prune_grace_period: 24h
+
 # Hostname to use in backup metadata
 # Default: system hostname
 #hostname: myserver

docs/DATAMODEL.md

@@ -1,273 +1,418 @@
-# Vaultik Data Model
+# vaultik data model
 
-## Overview
+`vaultik` keeps a SQLite database on the backed-up machine, called the local
+index. This document describes its tables and how each command uses them.
 
-Vaultik uses a local SQLite database to track file metadata, chunk mappings, and blob associations during the backup process. This database serves as an index for incremental backups and enables efficient deduplication.
+The local index has two jobs:
 
-**Important Notes:**
-- **No Migration Support (pre-1.0)**: Vaultik does not support database schema
-  migrations. The local index is treated as disposable — if the schema changes,
-  delete the local SQLite database (`vaultik database delete`) and run a full
-  backup. The remote storage is unaffected; the new index will re-deduplicate
-  against existing remote blobs.
-- **Version Compatibility**: In rare cases, you may need to use the same version
-  of Vaultik to restore a backup as was used to create it. This ensures
-  compatibility with the metadata format stored in S3.
+- It records which files, chunks and blobs are already stored. The next backup
+  then does not store them again.
+- It is the source of each snapshot's own database. A backup uploads a trimmed,
+  encrypted copy of the index with every snapshot.
 
-## Database Tables
+A restore does not need the local index. It reads the copy that was uploaded
+with the snapshot.
 
-### 1. `files`
-Stores metadata about files in the filesystem being backed up.
+## Terms
 
-**Columns:**
-- `id` (TEXT PRIMARY KEY) - UUID for the file record
-- `path` (TEXT NOT NULL UNIQUE) - Absolute file path
-- `mtime` (INTEGER NOT NULL) - Modification time as Unix timestamp
-- `size` (INTEGER NOT NULL) - File size in bytes
-- `mode` (INTEGER NOT NULL) - Unix file permissions and type
-- `uid` (INTEGER NOT NULL) - User ID of file owner
-- `gid` (INTEGER NOT NULL) - Group ID of file owner
-- `link_target` (TEXT) - Symlink target path (NULL for regular files)
+- **Snapshot**: one backup of the paths listed under one name in the config.
+- **Chunk**: a piece of a file's content, named by the SHA-256 hash of that
+  content.
+- **Blob**: one object at the destination. It holds many chunks, compressed and
+  then encrypted.
+- **Destination**: the place the backups are stored.
+- **Manifest**: the unencrypted list of the blobs one snapshot uses.
+- **Remote key**: a hash of the snapshot ID. It names the snapshot's directory
+  at the destination.
 
-**Indexes:**
-- `idx_files_path` on `path` for efficient lookups
+## Where the index is
 
-**Purpose:** Tracks file metadata to detect changes between backup runs. Used for incremental backup decisions. The UUID primary key provides stable references that don't change if files are moved.
+The index is the file `index.sqlite` in the platform data directory. That is
+`~/.local/share/vaultik/` on Linux and `~/Library/Application Support/vaultik/`
+on macOS. The `index_path` setting and the `VAULTIK_INDEX_PATH` environment
+variable change the path.
 
-### 2. `chunks`
-Stores information about content-defined chunks created from files.
+The file is readable only by its owner, because it lists every backed-up path.
+`vaultik` sets that mode each time it opens the index.
 
-**Columns:**
-- `chunk_hash` (TEXT PRIMARY KEY) - SHA256 hash of chunk content
-- `size` (INTEGER NOT NULL) - Chunk size in bytes
+## Schema migrations
 
-**Purpose:** Enables deduplication by tracking unique chunks across all files.
+`vaultik` supports schema migrations. They are the numbered SQL files in
+`internal/database/schema/`, which are built into the binary.
 
-### 3. `file_chunks`
-Maps files to their constituent chunks in order.
+- `000.sql` creates the `schema_migrations` table. That table records each
+  migration that has run.
+- `001.sql` creates every other table.
 
-**Columns:**
-- `file_id` (TEXT) - File ID (FK to files.id)
-- `idx` (INTEGER) - Chunk index within file (0-based)
-- `chunk_hash` (TEXT) - Chunk hash (FK to chunks.chunk_hash)
-- PRIMARY KEY (`file_id`, `idx`)
+`database.New` opens the index. It then applies, in order, every migration that
+the index has not recorded.
 
-**Purpose:** Allows reconstruction of files from chunks during restore.
+Before version 1.0, no migration is added, because nothing is installed anywhere
+yet. A schema change edits `001.sql` directly, with the code that uses the
+changed tables.
 
-### 4. `chunk_files`
-Reverse mapping showing which files contain each chunk.
+This has one consequence. An index created before such a change has already
+recorded `001.sql` as run, so it keeps the old schema and can become unusable.
+To rebuild it, run `vaultik database delete` and then a full backup.
 
-**Columns:**
-- `chunk_hash` (TEXT) - Chunk hash (FK to chunks.chunk_hash)
-- `file_id` (TEXT) - File ID (FK to files.id)
-- `file_offset` (INTEGER) - Byte offset of chunk within file
-- `length` (INTEGER) - Length of chunk in bytes
-- PRIMARY KEY (`chunk_hash`, `file_id`)
+After version 1.0, each schema change is a new numbered file. An existing index
+is then migrated the first time an updated `vaultik` opens it. A file that was
+part of a release is never edited.
 
-**Purpose:** Supports efficient queries for chunk usage and deduplication statistics.
+Migrations apply to the local index only. A snapshot's database at the
+destination is read as it was written. Each snapshot records the `vaultik`
+version that wrote it.
 
-### 5. `blobs`
-Stores information about packed, compressed, and encrypted blob files.
+## Tables
 
-**Columns:**
-- `id` (TEXT PRIMARY KEY) - UUID assigned when blob creation starts
-- `blob_hash` (TEXT UNIQUE) - SHA256 hash of final blob (NULL until finalized)
-- `created_ts` (INTEGER NOT NULL) - Creation timestamp
-- `finished_ts` (INTEGER) - Finalization timestamp (NULL if in progress)
-- `uncompressed_size` (INTEGER NOT NULL DEFAULT 0) - Total size of chunks before compression
-- `compressed_size` (INTEGER NOT NULL DEFAULT 0) - Size after compression and encryption
-- `uploaded_ts` (INTEGER) - Upload completion timestamp (NULL if not uploaded)
+Times are whole seconds since the Unix epoch, in UTC. The two exceptions are
+noted at their columns.
 
-**Purpose:** Tracks blob lifecycle from creation through upload. The UUID primary key allows immediate association of chunks with blobs.
+### `files`
 
-### 6. `blob_chunks`
-Maps chunks to the blobs that contain them.
+One row for each backed-up path: a regular file, a symlink or a directory.
 
-**Columns:**
-- `blob_id` (TEXT) - Blob ID (FK to blobs.id)
-- `chunk_hash` (TEXT) - Chunk hash (FK to chunks.chunk_hash)
-- `offset` (INTEGER) - Byte offset of chunk within blob (before compression)
-- `length` (INTEGER) - Length of chunk in bytes
-- PRIMARY KEY (`blob_id`, `chunk_hash`)
+- `id`: a UUID. Primary key.
+- `path`: the absolute path. Unique.
+- `source_path`: the configured path that the scan found this entry under. A
+  restore does not read it.
+- `mtime`: the modification time.
+- `mtime_nsec`: the nanoseconds within that second.
+- `size`: the size in bytes.
+- `mode`: the permission bits and the file type.
+- `uid`: the owner's user ID.
+- `gid`: the owner's group ID.
+- `link_target`: the target of a symlink. It is `NULL` for other entries.
 
-**Purpose:** Enables chunk retrieval from blobs during restore operations.
+`idx_files_path` indexes `path`.
 
-### 7. `snapshots`
-Tracks backup snapshots.
+There is one row for each path, and it is updated when the file changes. The
+row is shared by every snapshot in the local index that lists the path. An
+older snapshot's own view of the file is in that snapshot's database at the
+destination.
 
-**Columns:**
-- `id` (TEXT PRIMARY KEY) - Snapshot ID (format: hostname-YYYYMMDD-HHMMSSZ)
-- `hostname` (TEXT) - Hostname where backup was created
-- `vaultik_version` (TEXT) - Version of Vaultik used
-- `vaultik_git_revision` (TEXT) - Git revision of Vaultik used
-- `started_at` (INTEGER) - Start timestamp
-- `completed_at` (INTEGER) - Completion timestamp (NULL if in progress)
-- `file_count` (INTEGER) - Number of files in snapshot
-- `chunk_count` (INTEGER) - Number of unique chunks
-- `blob_count` (INTEGER) - Number of blobs referenced
-- `total_size` (INTEGER) - Total size of all files
-- `blob_size` (INTEGER) - Total size of all blobs (compressed)
-- `blob_uncompressed_size` (INTEGER) - Total uncompressed size of all referenced blobs
-- `compression_ratio` (REAL) - Compression ratio achieved
-- `compression_level` (INTEGER) - Compression level used for this snapshot
-- `upload_bytes` (INTEGER) - Total bytes uploaded during this snapshot
-- `upload_duration_ms` (INTEGER) - Total milliseconds spent uploading to S3
+### `chunks`
 
-**Purpose:** Provides snapshot metadata and statistics including version tracking for compatibility.
+One row for each distinct chunk.
 
-### 8. `snapshot_files`
-Maps snapshots to the files they contain.
+- `chunk_hash`: the SHA-256 hash of the chunk's content, in hex. Primary key.
+- `size`: the size in bytes.
 
-**Columns:**
-- `snapshot_id` (TEXT) - Snapshot ID (FK to snapshots.id)
-- `file_id` (TEXT) - File ID (FK to files.id)
-- PRIMARY KEY (`snapshot_id`, `file_id`)
+### `file_chunks`
 
-**Purpose:** Records which files are included in each snapshot.
+The chunks that make up each file, in order.
 
-### 9. `snapshot_blobs`
-Maps snapshots to the blobs they reference.
+- `file_id`: the file. References `files.id`.
+- `idx`: the position of the chunk within the file, counted from 0.
+- `chunk_hash`: the chunk. References `chunks.chunk_hash`.
 
-**Columns:**
-- `snapshot_id` (TEXT) - Snapshot ID (FK to snapshots.id)
-- `blob_id` (TEXT) - Blob ID (FK to blobs.id)
-- `blob_hash` (TEXT) - Denormalized blob hash for manifest generation
-- PRIMARY KEY (`snapshot_id`, `blob_id`)
+The primary key is (`file_id`, `idx`). `idx_file_chunks_chunk_hash` indexes
+`chunk_hash`.
 
-**Purpose:** Tracks blob dependencies for snapshots and enables manifest generation.
+### `chunk_files`
 
-### 10. `uploads`
-Tracks blob upload metrics.
+The reverse of `file_chunks`: the files that contain each chunk.
 
-**Columns:**
-- `blob_hash` (TEXT PRIMARY KEY) - Hash of uploaded blob
-- `snapshot_id` (TEXT NOT NULL) - The snapshot that triggered this upload (FK to snapshots.id)
-- `uploaded_at` (INTEGER) - Upload timestamp
-- `size` (INTEGER) - Size of uploaded blob
-- `duration_ms` (INTEGER) - Upload duration in milliseconds
+- `chunk_hash`: the chunk. References `chunks.chunk_hash`.
+- `file_id`: the file. References `files.id`.
+- `file_offset`: the byte offset of the chunk within the file.
+- `length`: the length of the chunk in bytes.
 
-**Purpose:** Performance monitoring and tracking which blobs were newly created (uploaded) during each snapshot.
+The primary key is (`chunk_hash`, `file_id`). `idx_chunk_files_file_id` indexes
+`file_id`.
 
-## Data Flow and Operations
+### `blobs`
 
-### 1. Backup Process
+One row for each blob.
 
-1. **File Scanning**
-   - `INSERT OR REPLACE INTO files` - Update file metadata
-   - `SELECT * FROM files WHERE path = ?` - Check if file has changed
-   - `INSERT INTO snapshot_files` - Add file to current snapshot
+- `id`: a UUID, given when packing of the blob starts. Primary key.
+- `blob_hash`: the name of the blob at the destination. Unique. Until the blob
+  is closed, it holds a placeholder that starts with `temp-placeholder-`.
+- `created_ts`: when packing started.
+- `finished_ts`: when the blob was closed. It is `NULL` before that.
+- `uncompressed_size`: the total size of the blob's chunks.
+- `compressed_size`: the size after compression and encryption.
+- `uploaded_ts`: when the upload finished. It is `NULL` before that.
 
-2. **Chunking** (for changed files)
-   - `INSERT OR IGNORE INTO chunks` - Store new chunks
-   - `INSERT INTO file_chunks` - Map chunks to file
-   - `INSERT INTO chunk_files` - Create reverse mapping
+`blob_hash` is `hex(SHA256(SHA256(uncompressed contents)))`. See
+[REPOSTRUCTURE.md](REPOSTRUCTURE.md#blobs).
 
-3. **Blob Packing**
-   - `INSERT INTO blobs` - Create blob record with UUID (blob_hash NULL)
-   - `INSERT INTO blob_chunks` - Associate chunks with blob immediately
-   - `UPDATE blobs SET blob_hash = ?, finished_ts = ?` - Finalize blob after packing
+Once a blob is uploaded, `compressed_size` is the size of the object at the
+destination. A blob that was already there can have a different size from the
+local copy. That happens when it was written with another compression level or
+other recipients.
 
-4. **Upload**
-   - `UPDATE blobs SET uploaded_ts = ?` - Mark blob as uploaded
-   - `INSERT INTO uploads` - Record upload metrics with snapshot_id
-   - `INSERT INTO snapshot_blobs` - Associate blob with snapshot
+### `blob_chunks`
 
-5. **Snapshot Completion**
-   - `UPDATE snapshots SET completed_at = ?, stats...` - Finalize snapshot
-   - Generate and upload blob manifest from `snapshot_blobs`
+The chunks that each blob holds.
 
-### 2. Incremental Backup
+- `blob_id`: the blob. References `blobs.id`.
+- `chunk_hash`: the chunk. References `chunks.chunk_hash`.
+- `offset`: the byte offset of the chunk within the uncompressed blob.
+- `length`: the length of the chunk in bytes.
 
-1. **Change Detection**
-   - `SELECT * FROM files WHERE path = ?` - Get previous file metadata
-   - Compare mtime, size, mode to detect changes
-   - Skip unchanged files but still add to `snapshot_files`
+The primary key is (`blob_id`, `chunk_hash`). `idx_blob_chunks_chunk_hash`
+indexes `chunk_hash`.
 
-2. **Chunk Reuse**
-   - `SELECT * FROM blob_chunks WHERE chunk_hash = ?` - Find existing chunks
-   - `INSERT INTO snapshot_blobs` - Reference existing blobs for unchanged files
+A restore finds the blob for each chunk through this table.
 
-### 3. Snapshot Metadata Export
+### `snapshots`
 
-After a snapshot is completed:
-1. Copy database to temporary file
-2. Clean temporary database to contain only current snapshot data
-3. Export to SQL dump using sqlite3
-4. Compress with zstd and encrypt with age
-5. Upload to S3 as `metadata/{snapshot-id}/db.zst.age`
-6. Generate blob manifest and upload as `metadata/{snapshot-id}/manifest.json.zst`
+One row for each snapshot.
 
-### 4. Restore Process
+- `id`: the snapshot ID. Primary key.
+- `hostname`: the hostname of the machine that made the backup.
+- `vaultik_version`: the `vaultik` version that made it.
+- `vaultik_git_revision`: the git commit of that version.
+- `started_at`: when the backup started.
+- `completed_at`: when the backup finished. It is `NULL` until then.
+- `file_count`: the number of files and symlinks in the snapshot. Directories
+  are not counted.
+- `chunk_count`: the number of chunks this snapshot stored that were not stored
+  before.
+- `new_chunk_size`: the total size of those chunks before compression.
+- `blob_count`: the number of blobs this snapshot created.
+- `total_size`: the total size of all files in the snapshot.
+- `uncompressed_chunk_size`: the total size, before compression, of the
+  distinct chunks the snapshot's files use.
+- `blob_size`: the total compressed size of the blobs the snapshot uses.
+- `blob_uncompressed_size`: the total uncompressed size of those blobs.
+- `compression_ratio`: the compression ratio of those blobs.
+- `compression_level`: the compression level the backup used.
+- `upload_bytes`: the bytes uploaded during the backup.
+- `upload_duration_ms`: the time spent uploading, in milliseconds.
 
-The restore process doesn't use the local database. Instead:
-1. Downloads snapshot metadata from S3
-2. Downloads required blobs based on manifest
-3. Reconstructs files from decrypted and decompressed chunks
+The snapshot ID has the form `<hostname>_<snapshot-name>_<timestamp>`. An
+example is `server1_home_2025-06-01T12:00:00Z`. It has three parts:
 
-### 5. Pruning
+- the hostname, up to its first `.`
+- the snapshot name
+- the start time, in UTC
 
-1. **Identify Unreferenced Blobs**
-   - Query blobs not referenced by any remaining snapshot
-   - Delete from S3 and local database
+`snapshot list` shows `uncompressed_chunk_size` as `UNCOMPRESSED SIZE` and
+`new_chunk_size` as `NEW CHUNK SIZE`.
 
-### 6. Incomplete Snapshot Cleanup
+A snapshot with `completed_at` set to `NULL` is unfinished. `snapshot list`
+does not show it. The next backup deletes it, and so does `prune`.
 
-Before each backup:
-1. Query incomplete snapshots (where `completed_at IS NULL`)
-2. Check if metadata exists in S3
-3. If no metadata, delete snapshot and all associations
-4. Clean up orphaned files, chunks, and blobs
+### `snapshot_files`
 
-## Repository Pattern
+The files in each snapshot.
 
-Vaultik uses a repository pattern for database access:
+- `snapshot_id`: the snapshot. References `snapshots.id`.
+- `file_id`: the file. References `files.id`.
 
-- `FileRepository` - CRUD operations for files and file metadata
-- `ChunkRepository` - CRUD operations for content chunks
-- `FileChunkRepository` - Manage file-to-chunk mappings
-- `ChunkFileRepository` - Manage chunk-to-file reverse mappings  
-- `BlobRepository` - Manage blob lifecycle (creation, finalization, upload)
-- `BlobChunkRepository` - Manage blob-to-chunk associations
-- `SnapshotRepository` - Manage snapshots and their relationships
-- `UploadRepository` - Track blob upload metrics
+The primary key is (`snapshot_id`, `file_id`). `idx_snapshot_files_file_id`
+indexes `file_id`.
 
-Each repository provides methods like:
-- `Create()` - Insert new record
-- `GetByID()` / `GetByPath()` / `GetByHash()` - Retrieve records
-- `Update()` - Update existing records
-- `Delete()` - Remove records
-- Specialized queries for each entity type (e.g., `DeleteOrphaned()`, `GetIncompleteByHostname()`)
+### `snapshot_blobs`
 
-## Transaction Management
+The blobs that each snapshot uses.
 
-All database operations that modify multiple tables are wrapped in transactions:
+- `snapshot_id`: the snapshot. References `snapshots.id`.
+- `blob_id`: the blob. References `blobs.id`.
+- `blob_hash`: a copy of the blob's hash.
+
+The primary key is (`snapshot_id`, `blob_id`). `idx_snapshot_blobs_blob_id`
+indexes `blob_id`.
+
+The manifest is built from this table.
+
+### `snapshot_mount_points`
+
+The mount points that a snapshot's backup found.
+
+- `snapshot_id`: the snapshot. References `snapshots.id`.
+- `path`: the absolute path of the directory.
+
+The primary key is (`snapshot_id`, `path`).
+
+A directory is recorded here when all of these hold:
+
+- It is at or below a configured path.
+- The backup found an entry in it that is not excluded.
+- It is on a different filesystem from its parent directory.
+
+The device number is read after the directory has been listed. Listing an
+automount point mounts its volume, so the number is that of the volume. An
+automount point whose volume cannot be mounted is not recorded, because the
+backup finds no entry in it.
+
+`snapshot create --prune` reads this table. It skips the prune when an earlier
+snapshot of the same name recorded a mount point that the new backup did not
+find. That happens when the volume is not mounted.
+
+### `uploads`
+
+One row for each blob that a backup uploaded.
+
+- `blob_hash`: the blob. References `blobs.blob_hash`. Primary key.
+- `snapshot_id`: the snapshot whose backup uploaded it. References
+  `snapshots.id`.
+- `uploaded_at`: when the upload started. It is written as a date and time, not
+  as seconds.
+- `size`: the size of the uploaded blob.
+- `duration_ms`: the duration of the upload, in milliseconds.
+
+`idx_uploads_snapshot_id` indexes `snapshot_id`.
+
+A blob that was already at the destination gets no row here.
+
+### `local_meta`
+
+Settings that belong to this copy of the index.
+
+- `key`: the name of the setting. Primary key.
+- `value`: its value.
+
+The only key is `storage_url`. It holds the destination that the index was
+first used with. A command that changes data refuses to run when the config
+names a different destination.
+
+### `schema_migrations`
+
+The migrations that have run.
+
+- `version`: the number of the migration. Primary key.
+- `applied_at`: when it ran. It is written as a date and time, not as seconds.
+
+## How commands use the tables
+
+### Backup
+
+A backup works on the tables in this order:
+
+- It deletes unfinished snapshots, and the rows that no snapshot uses.
+- It creates a row in `snapshots`, with `completed_at` set to `NULL`.
+- It deletes the `blobs` rows that were never uploaded.
+- It walks each path and compares every entry with its row in `files`.
+- It links each unchanged file to the snapshot in `snapshot_files`.
+- It records each mount point in `snapshot_mount_points`.
+- It reads each changed file and stores it. See below.
+- It fills `snapshot_blobs` with every blob the snapshot uses.
+- It stores the snapshot's totals in `snapshots`.
+- It uploads the snapshot's metadata.
+- It sets `completed_at`.
+
+A file counts as changed when its row in `files` differs from the file on disk
+in any of these:
+
+- size
+- modification time
+- mode
+- owner or group
+- link target
+
+A file is also read again when it lists a chunk that no uploaded blob holds.
+
+Storing a changed file writes these rows:
+
+- A row in `blobs`, when packing of a blob starts.
+- When the blob is closed, in one transaction:
+    - a row in `chunks` for each new chunk in the blob
+    - the blob's rows in `blob_chunks`
+    - the blob's `blob_hash`, sizes and `finished_ts`
+- When the blob is stored at the destination, in one transaction:
+    - `uploaded_ts` and `compressed_size` in `blobs`
+    - a row in `snapshot_blobs`
+    - a row in `uploads`, unless the blob was already at the destination
+- The row in `files`, with its `file_chunks`, `chunk_files` and
+  `snapshot_files` rows.
+
+The row in `files` is written last. It is written only after every blob that
+holds the file's new chunks is stored.
+
+### Snapshot metadata
+
+The metadata of a snapshot is made from a copy of the index:
+
+- The index is copied into a temporary directory with `VACUUM INTO`.
+- The rows of every other snapshot are deleted from the copy. So are the files
+  and blobs that this snapshot does not use. A chunk is then deleted when no
+  remaining file and no remaining blob uses it.
+- The `local_meta` and `schema_migrations` tables are kept as they are. The
+  uploaded database therefore names the destination, and lists the migrations
+  that have run.
+- `VACUUM` rewrites the copy, so that no deleted row is left in the file.
+- The copy is compressed with zstd and encrypted with age.
+- The manifest is built from the copy's `snapshot_blobs` table.
+- The copy is uploaded as `metadata/<remote-key>/db.zst.age`.
+- The manifest is uploaded as `metadata/<remote-key>/manifest.json.zst`.
+
+The copy is made through SQLite, not by copying the file. The index runs in WAL
+mode, so rows that are already committed can still be in the `-wal` file beside
+it.
+
+### Restore
+
+A restore does not use the local index.
+
+- It downloads and decrypts `db.zst.age` into a temporary directory.
+- It opens that file read-only. It applies no migration to it.
+- It reads the files to restore from `files`.
+- It finds each file's chunks in `file_chunks`.
+- It finds the blob and offset of each chunk in `blob_chunks`.
+- It finds the name of each blob in `blobs`.
+
+It does not read the manifest.
+
+### Prune
+
+`prune` changes the local index in two ways:
+
+- It deletes each snapshot whose manifest is gone from the destination.
+- It deletes unfinished snapshots, and the rows that no snapshot uses.
+
+The second step removes unused rows in this order:
+
+- files that are in no snapshot
+- blobs that are in no snapshot
+- `blob_chunks` rows whose chunk is gone
+- chunks that no file and no blob uses
+
+`prune` then deletes unused data at the destination. That step reads the
+manifests, not the local index. It reads the local index only to keep a
+directory under `metadata/` whose snapshot is still recorded there.
+
+## Code
+
+Each repository type in `internal/database` serves one or more tables:
+
+- `FileRepository`: `files`
+- `ChunkRepository`: `chunks`
+- `FileChunkRepository`: `file_chunks`
+- `ChunkFileRepository`: `chunk_files`
+- `BlobRepository`: `blobs`
+- `BlobChunkRepository`: `blob_chunks`
+- `SnapshotRepository`:
+    - `snapshots`
+    - `snapshot_files`
+    - `snapshot_blobs`
+    - `snapshot_mount_points`
+- `UploadRepository`: `uploads`
+- `LocalMetaRepository`: `local_meta`
+
+No repository type serves `schema_migrations`. The code that applies the
+migrations reads and writes it.
+
+`Repositories` holds one of each type.
+
+`Repositories.WithTx` runs a function in one transaction:
 
 ```go
 err := repos.WithTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
-    // Multiple repository operations using tx
+    // several repository calls that use tx
 })
 ```
 
-This ensures consistency, especially important for operations like:
-- Creating file-chunk mappings
-- Associating chunks with blobs
-- Updating snapshot statistics
+The transaction commits when the function returns no error. Otherwise it is
+rolled back.
 
-## Performance Considerations
+## SQLite settings
 
-1. **Indexes**: 
-   - Primary keys are automatically indexed
-   - `idx_files_path` on `files(path)` for efficient file lookups
-
-2. **Prepared Statements**: All queries use prepared statements for performance and security
-
-3. **Batch Operations**: Where possible, operations are batched within transactions
-
-4. **Write-Ahead Logging**: SQLite WAL mode is enabled for better concurrency
-
-## Data Integrity
-
-1. **Foreign Keys**: Enforced through CASCADE DELETE and application-level repository methods
-2. **Unique Constraints**: Chunk hashes, file paths, and blob hashes are unique
-3. **Null Handling**: Nullable fields clearly indicate in-progress operations
-4. **Timestamp Tracking**: All major operations record timestamps for auditing
+- The index runs in WAL mode. A command that only reads, such as
+  `snapshot list`, can read the index while a backup writes to it.
+- A connection waits up to 10 seconds for a lock before it fails.
+- Foreign keys are enforced.
+- Each open database uses one connection, so its queries run one at a time.
+- Queries pass their values as parameters.

docs/REPOSTRUCTURE.md

@@ -1,144 +1,416 @@
-# Vaultik S3 Repository Structure
+# vaultik destination layout
 
-This document describes the structure and organization of data stored in the S3 bucket by Vaultik.
+This document describes what `vaultik` stores at the destination. It covers the
+layout, what the layout protects, and the risks it accepts.
 
-## Overview
+The destination is the place the backups are stored. It is an S3 bucket and
+prefix, a local directory or an rclone remote. The layout is the same on all
+three.
 
-Vaultik stores all backup data in an S3-compatible object store. The repository consists of two main components:
-1. **Blobs** - The actual backup data (content-addressed, encrypted)
-2. **Metadata** - Snapshot information and manifests (partially encrypted)
+## Terms
 
-## Directory Structure
+- **Snapshot**: one backup of the paths listed under one name in the config.
+- **Snapshot ID**: the name of a snapshot, in the form
+  `<hostname>_<snapshot-name>_<timestamp>`.
+- **Chunk**: a piece of a file's content, named by the SHA-256 hash of that
+  content.
+- **Blob**: one object under `blobs/`. It holds many chunks.
+- **Local index**: the SQLite database on the backed-up machine.
+- **Recipient**: an age public key that backups are encrypted to.
+
+## Layout
 
 ```
-<bucket>/<prefix>/
+<destination>/
+├── vaultik-destination.txt
 ├── blobs/
 │   └── <hash[0:2]>/
 │       └── <hash[2:4]>/
-│           └── <full-hash>
+│           └── <hash>
 └── metadata/
-    └── <snapshot-id>/
+    └── <remote-key>/
         ├── db.zst.age
         └── manifest.json.zst
 ```
 
-## Blobs Directory (`blobs/`)
+The destination holds three things:
 
-### Structure
-- **Path format**: `blobs/<first-2-chars>/<next-2-chars>/<full-hash>`
-- **Example**: `blobs/ca/fe/cafebabe1234567890abcdef1234567890abcdef1234567890abcdef12345678`
-- **Sharding**: The two-level directory structure (using the first 4 characters of the hash) prevents any single directory from containing too many objects
+- `vaultik-destination.txt`, the marker file
+- `blobs/`, which holds the file contents
+- `metadata/`, which holds one directory for each snapshot
 
-### Content
-- **What it contains**: Packed collections of content-defined chunks from files
-- **Format**: Zstandard compressed, then Age encrypted
-- **Encryption**: Always encrypted with Age using the configured recipients
-- **Naming**: Content-addressed using SHA256 hash of the encrypted blob
+## The marker file
 
-### Why Encrypted
-Blobs contain the actual file data from backups and must be encrypted for security. The content-addressing ensures deduplication while the encryption ensures privacy.
+Every backup writes `vaultik-destination.txt` before it uploads anything else.
+The file holds one fixed line of text. It names no host, path or time.
 
-## Metadata Directory (`metadata/`)
+The marker tells an emptied or missing destination apart from a new one. An
+unplugged volume leaves an empty directory at its mount point. Without the
+marker, that directory would look like a destination whose snapshots are all
+gone.
 
-Each snapshot has its own subdirectory named with the snapshot ID.
+The check applies while the local index holds snapshots. When the destination
+then answers that the marker does not exist, commands treat the destination as
+not present:
 
-### Snapshot ID Format
-- **Format**: `<hostname>_<snapshot-name>_<RFC3339>` (or `<hostname>_<RFC3339>` if no
-  name was specified)
-- **Example**: `laptop_home_2024-01-15T14:30:52Z`
-- **Components**:
-  - Short hostname (everything before the first dot is stripped from the FQDN)
-  - Snapshot name from the configured `snapshots:` map (optional)
-  - RFC3339 UTC timestamp
+- `snapshot create` refuses to write to it.
+- `prune`, `snapshot purge` and `remote nuke` fail.
+- `snapshot list` shows the local index alone.
+- `snapshot remove` deletes nothing at the destination.
 
-### Files in Each Snapshot Directory
+The check is skipped when the look-up fails for any other reason. S3 refuses
+the look-up to credentials that can only write. A machine with such credentials
+therefore does not notice an emptied destination. Its next backup would not
+upload data that its local index says is stored. After a destination is emptied,
+run `vaultik database delete` on such a machine before it backs up again.
 
-#### `db.zst.age` - Encrypted Database
-- **What it contains**: Pruned binary SQLite database for this snapshot
-- **Format**: Binary SQLite → Zstandard compressed → Age encrypted
-- **Encryption**: Encrypted with Age
-- **Purpose**: Contains full file metadata, chunk mappings, and all relationships
-- **Why encrypted**: Contains sensitive metadata like file paths, permissions, and ownership
+`remote nuke` deletes the marker last.
+
+## Blobs
+
+A blob holds the content of many chunks, one after another.
+
+- **Path**: `blobs/<first 2 characters>/<next 2 characters>/<full hash>`.
+- **Example**:
+  `blobs/ca/fe/cafebabe1234567890abcdef1234567890abcdef1234567890abcdef12345678`.
+- **Format**: the chunk data, compressed with zstd and then encrypted with age.
+- **Recipients**: the public keys in `age_recipients` when the blob was written.
+
+The two directory levels keep any one directory from holding too many objects.
+
+### Blob names
+
+A blob's name is `hex(SHA256(SHA256(uncompressed contents)))`. The contents are
+the chunk data, joined in order, before compression and encryption. The name is
+not a hash of the stored bytes.
+
+This has two consequences:
+
+- Checking a blob against its name needs the private key. The blob must be
+  decrypted and decompressed first. `snapshot restore` and
+  `snapshot verify --deep` do this.
+- Without the key, only a blob's presence and size can be checked.
+  `snapshot verify` does this.
+
+`blobgen.DoubleSHA256` in `internal/blobgen` computes the name.
+
+### Uploads
+
+A backup looks a blob up by name before it uploads it. It does not upload a
+blob that is already there.
+
+When the destination refuses the look-up, the backup uploads the blob. An
+object with that name is then replaced by one that holds the same chunk data.
+
+## Metadata
+
+Each snapshot has one directory under `metadata/`. The directory is named with
+the snapshot's remote key, not with its snapshot ID.
+
+### Snapshot IDs
+
+A snapshot ID has the form `<hostname>_<snapshot-name>_<timestamp>`. An example
+is `laptop_home_2024-01-15T14:30:52Z`. It has three parts:
+
+- the hostname, up to its first `.`
+- the snapshot name from the config
+- the time the backup started, in UTC
+
+The snapshot ID is used on the command line, in log records and in the local
+index. It reveals the hostname, the snapshot name and the backup time. It is
+therefore never written to the destination without encryption.
+
+### Remote keys
+
+The remote key is `hex(SHA256(SHA256("vaultik|" + snapshot ID)))`. It is 64 hex
+characters long. `internal/snapshot/remotekey.go` computes it.
+
+For example:
+
+- Snapshot ID: `server1_home_2025-06-01T12:00:00Z`
+- Remote key: `17f97bcde958748af076b926af59823943db59e80ce7170b40f124dfa28f64aa`
+- Directory:
+  `metadata/17f97bcde958748af076b926af59823943db59e80ce7170b40f124dfa28f64aa/`
+
+A listing of the destination therefore shows no hostname and no snapshot name.
+
+The hash cannot be reversed, but it uses no secret. Someone who guesses a
+snapshot ID can hash it the same way and see whether that snapshot exists. The
+remote key keeps names out of a listing. It does not hide them from a guess.
+
+### `db.zst.age`
+
+The snapshot's database. A restore needs it.
+
+- **Contents**: a SQLite database. It is a copy of the local index without the
+  rows of other snapshots. It lists every file in the snapshot with its
+  metadata, its chunks, and the blob that holds each chunk.
+- **Format**: the database file, compressed with zstd and then encrypted with
+  age.
+
+It is encrypted because it holds file paths, permissions and owners.
+[DATAMODEL.md](DATAMODEL.md) describes its tables.
+
+### `manifest.json.zst`
+
+The list of the blobs the snapshot uses.
+
+- **Format**: JSON, compressed with zstd. It is not encrypted.
 
-#### `manifest.json.zst` - Unencrypted Blob Manifest
-- **What it contains**: JSON list of all blob hashes referenced by this snapshot
-- **Format**: JSON → Zstandard compressed (NOT encrypted)
-- **Encryption**: NOT encrypted
-- **Purpose**: Enables pruning operations without requiring decryption keys
-- **Structure**:
 ```json
 {
-  "snapshot_id": "laptop_home_2024-01-15T14:30:52Z",
-  "timestamp": "2024-01-15T14:30:52Z",
-  "blob_count": 42,
+    "snapshot_id": "17f97bcde958748af076b926af59823943db59e80ce7170b40f124dfa28f64aa",
+    "timestamp": "2025-06-01T12:04:31Z",
+    "blob_count": 2,
+    "total_compressed_size": 57344,
     "blobs": [
-    "cafebabe1234567890abcdef1234567890abcdef1234567890abcdef12345678",
-    "deadbeef1234567890abcdef1234567890abcdef1234567890abcdef12345678",
-    ...
+        {
+            "hash": "cafebabe1234567890abcdef1234567890abcdef1234567890abcdef12345678",
+            "compressed_size": 24576
+        },
+        {
+            "hash": "deadbeef1234567890abcdef1234567890abcdef1234567890abcdef12345678",
+            "compressed_size": 32768
+        }
     ]
 }
 ```
 
-### Why Manifest is Unencrypted
-The manifest must be readable without the private key to enable:
-1. **Pruning operations** - Identifying unreferenced blobs for deletion
-2. **Storage analysis** - Understanding space usage without decryption
-3. **Verification** - Checking blob existence without decryption
-4. **Cross-snapshot deduplication analysis** - Finding shared blobs between snapshots
+- `snapshot_id`: the remote key, not the snapshot ID.
+- `timestamp`: when the manifest was written, at the end of the backup.
+- `blob_count`: the number of blobs.
+- `total_compressed_size`: the total size of those blobs, in bytes.
+- `blobs`: the hash and the size of each blob.
 
-The manifest only contains blob hashes, not file names or any other sensitive information.
+The manifest holds no file name and no path.
 
-## Security Considerations
+The manifest is not encrypted so that these commands work without the private
+key:
 
-### What's Encrypted
-- **All file content** (in blobs)
-- **All file metadata** (paths, permissions, timestamps, ownership in db.zst.age)
-- **File-to-chunk mappings** (in db.zst.age)
+- `prune`, which must know which blobs are still in use
+- `snapshot verify`, which checks that each blob is present
+- `snapshot list` and `remote info`, which report what the destination holds
 
-### What's Not Encrypted
-- **Blob hashes** (in manifest.json.zst)
-- **Snapshot IDs** (directory names)
-- **Blob count per snapshot** (in manifest.json.zst)
+## What is protected
 
-### Privacy Implications
-From the unencrypted data, an observer can determine:
-- When backups were taken (from snapshot IDs)
-- Which hostname created backups (from snapshot IDs)
-- How many blobs each snapshot references
-- Which blobs are shared between snapshots (deduplication patterns)
-- The size of each encrypted blob
+These are encrypted:
 
-An observer cannot determine:
-- File names or paths
-- File contents
-- File permissions or ownership
-- Directory structure
-- Which chunks belong to which files
+- file contents, in the blobs
+- in `db.zst.age`:
+    - file names and paths
+    - permissions and owners
+    - modification times
+    - which chunks make up which file
 
-## Consistency Guarantees
+These are not encrypted:
 
-1. **Blobs are immutable** - Once written, a blob is never modified
-2. **Blobs are written before metadata** - A snapshot's metadata is only written after all its blobs are successfully uploaded
-3. **Metadata is written atomically** - Both db.zst.age and manifest.json.zst are written as complete files
-4. **Snapshots are marked complete in local DB only after metadata upload** - Ensures consistency between local and remote state
+- the remote key of each snapshot
+- the time each backup finished, in the manifest
+- the hash and the compressed size of each blob, in the manifest
+- the number of blobs and their total size for each snapshot, in the manifest
 
-## Pruning Safety
+Someone who can read the destination therefore learns these:
 
-The prune operation is safe because:
-1. It only deletes blobs not referenced in any manifest
-2. Manifests are unencrypted and can be read without keys
-3. The operation compares the latest local DB snapshot with the latest S3 snapshot to ensure consistency
-4. Pruning will fail if these don't match, preventing accidental deletion of needed blobs
+- when each backup finished
+- how many blobs each snapshot uses, and how large they are
+- which blobs two snapshots share
+- whether a snapshot ID they guess exists
 
-## Restoration Requirements
+They do not learn these:
 
-To restore from a backup, you need:
-1. **The Age private key** - To decrypt blobs and database
-2. **The snapshot metadata** - Both files from the snapshot's metadata directory
-3. **All referenced blobs** - As listed in the manifest
+- any hostname or snapshot name, unless they guess it
+- file names or paths
+- file contents
+- permissions or owners
+- the directory structure
+- which chunks belong to which file
 
-The restoration process:
-1. Download and decrypt the database dump to understand file structure
-2. Download and decrypt the required blobs
-3. Reconstruct files from their chunks
-4. Restore file metadata (permissions, timestamps, etc.)
+The times and sizes could not be hidden by encrypting the manifest. The
+destination itself shows when each object was written and how large it is.
+
+## Accepted risks
+
+These are known properties of the format. They are written down so that you can
+weigh them.
+
+### No proof of who wrote the data
+
+A restore and `snapshot verify --deep` prove two things. The data decrypts with
+your private key, and it matches its hashes. They do not prove who wrote it.
+
+The hashes use no secret. Someone who knows a recipient's public key and can
+replace objects at the destination can put a snapshot of their own there. The
+public key is not stored at the destination, but a backed-up machine that is
+broken into has it.
+
+The defences are on the destination's side:
+
+- object versioning or object lock on the bucket
+- credentials for the backed-up machine that cannot delete or overwrite
+- running `prune` from a trusted machine
+
+An S3 `PutObject` replaces an existing object. Permission to write alone
+therefore does not stop overwrites.
+
+### Sizes are visible
+
+An object's size follows exactly from the size of its compressed contents.
+
+All the new chunks in one blob share one zstd stream. The stream looks back 8
+MiB, or 4 MiB at compression levels 1 and 2. A blob is closed at
+`blob_size_limit` and at the end of each configured path. A chunk that is
+already stored is never packed again.
+
+Data that an outsider writes compresses smaller when it matches a secret in the
+same chunk, so the blob gets smaller. The blob's size therefore shows the
+outsider whether their data matched the secret.
+
+This works only when all of these hold:
+
+- The outsider can write into a backed-up file.
+- The outsider can see the blob sizes at the destination.
+- The outsider's data and the secret land in the same chunk.
+- The file keeps changing.
+
+Back up a directory that outsiders can write to as a snapshot of its own.
+
+### Chunk boundaries use no secret
+
+The FastCDC parameters are fixed and public. `chunk_size` sets the average chunk
+size. No chunk is larger than four times this size. Only a file's last chunk can
+be smaller than a quarter of it. With the default of 10 MB, those limits are 40
+MB and 2.5 MB. A file of 2.5 MB or less is one chunk.
+
+At the default `blob_size_limit` of 10 GB, a blob holds hundreds of chunks. The
+size of the blob then does not show the size of any one chunk. A limit close to
+`chunk_size` begins to show chunk sizes.
+
+### Decrypted data on the local disk
+
+These commands write unencrypted data to the temporary directory:
+
+- `snapshot restore` writes decrypted blobs under `vaultik-blobcache-*/`. There
+  is no limit on their total size.
+- `snapshot restore` writes the decrypted database to
+  `vaultik-restore-*/snapshot.db`.
+- `snapshot verify --deep` writes the decrypted database to
+  `vaultik-verify-*/snapshot.db`.
+- `snapshot create` writes a copy of the local index to
+  `vaultik-snapshot-*/snapshot.db`.
+
+These files are readable only by their owner, and are removed when the command
+ends. A `kill -9` or a power loss leaves them behind. Delete any leftover
+`vaultik-*` entry in the temporary directory yourself.
+
+Trust the temporary directory as much as the directory you restore into.
+
+### Commands need different permissions
+
+`snapshot create` needs only permission to write objects. Other commands need
+more:
+
+- `snapshot list`, `snapshot verify`, `snapshot restore` and `remote info` list
+  and read.
+- `snapshot purge`, `snapshot remove` and `remote nuke` list and delete.
+- `prune` lists, reads and deletes.
+
+`snapshot create --prune` runs a prune on the backed-up machine. That machine
+then holds credentials that can delete its own backups. To keep it to
+write-only credentials, run `prune` from a separate, trusted machine.
+
+### New recipients cannot read old data
+
+Changing `age_recipients` does not re-encrypt what is already stored. A chunk
+that is already stored is not written again, so it stays encrypted to the
+earlier recipients.
+
+A new snapshot that uses such chunks cannot be restored with a new recipient's
+key alone. To make everything readable by a new key, run
+`vaultik database delete` and take a full backup to a new destination.
+
+### X25519 recipients only
+
+`vaultik` accepts age X25519 public keys only. It rejects age SSH and plugin
+recipients.
+
+The encrypted data is held by a third party for a long time. There is no second
+kind of recipient to fall back on if X25519 is broken, and no post-quantum
+option.
+
+## Order of writes
+
+A backup writes in a fixed order:
+
+- the marker file
+- the blobs
+- `db.zst.age`
+- `manifest.json.zst`
+
+A snapshot's metadata is written only after all its blobs are uploaded. The
+manifest is written last. A directory under `metadata/` that has no manifest
+therefore belongs to a backup that did not finish.
+
+The local index marks a snapshot complete only after both metadata files are
+uploaded. A backup that stops during that upload leaves the snapshot
+unfinished. The next backup removes it from the local index and stores its data
+again.
+
+One short gap remains. A backup can stop after the upload and before the
+snapshot is marked complete. The destination then holds a snapshot that can be
+restored, and the local index does not list it. `snapshot list` shows it as
+remote only.
+
+An upload that is cut off leaves no partial object under the final name on S3
+or on a local directory. The same holds for an rclone remote that can move
+objects. An rclone remote that cannot move objects can keep a partial object.
+
+## Pruning
+
+`prune` deletes two kinds of object at the destination:
+
+- a blob that no manifest lists
+- a directory under `metadata/` that has no manifest
+
+It reads the manifests, which are not encrypted, so it needs no private key.
+
+These rules keep it from deleting data that is in use:
+
+- It keeps every blob that any manifest lists.
+- It deletes nothing when any manifest cannot be downloaded or decoded.
+- It deletes a blob only once the blob is older than `prune_grace_period`.
+- It deletes a directory only once every object in it is that old.
+- It keeps a directory whose snapshot is in the local index.
+- It leaves alone a directory whose name is not a remote key.
+
+`prune_grace_period` defaults to `24h`. It protects a backup that is still
+running. Such a backup has uploaded blobs and perhaps `db.zst.age`, but not yet
+its manifest. A prune from this machine or another keeps those objects, as long
+as no backup takes longer than the grace period.
+
+The grace period does not protect one case. A backup does not upload a blob
+that is already at the destination, so that blob keeps the time of its first
+upload. The blob can be left over from an earlier backup that stopped before
+its manifest. A prune that runs before the new manifest is written can then
+delete that blob. The new snapshot cannot be restored afterwards.
+
+To avoid this case, give each machine and each user its own destination.
+
+`remote nuke` applies none of these rules.
+
+## What a restore needs
+
+- the age private key
+- the snapshot's `db.zst.age`
+- the blobs that hold the files to restore
+
+A restore does not read the manifest and does not need the local index.
+
+It works in these steps:
+
+- It downloads and decrypts `db.zst.age`.
+- It checks that the database is the one for the snapshot that was asked for.
+- It downloads and decrypts the blobs it needs. It checks each blob against its
+  name.
+- It writes each file from its chunks.
+- It sets each file's permissions and times, and its owner when it runs as
+  root.

go.mod

@@ -20,9 +20,12 @@ require (
 	github.com/rclone/rclone v1.72.1
 	github.com/spf13/afero v1.15.0
 	github.com/spf13/cobra v1.10.1
+	github.com/spf13/pflag v1.0.10
 	github.com/stretchr/testify v1.11.1
+	go.uber.org/dig v1.19.0
 	go.uber.org/fx v1.24.0
 	golang.org/x/sync v0.18.0
+	golang.org/x/sys v0.38.0
 	golang.org/x/term v0.37.0
 	gopkg.in/yaml.v3 v3.0.1
 	modernc.org/sqlite v1.38.0
@@ -225,7 +228,6 @@ require (
 	github.com/smarty/assertions v1.16.0 // indirect
 	github.com/sony/gobreaker v1.0.0 // indirect
 	github.com/spacemonkeygo/monkit/v3 v3.0.25-0.20251022131615-eb24eb109368 // indirect
-	github.com/spf13/pflag v1.0.10 // indirect
 	github.com/t3rm1n4l/go-mega v0.0.0-20251031123324-a804aaa87491 // indirect
 	github.com/tidwall/gjson v1.18.0 // indirect
 	github.com/tidwall/match v1.1.1 // indirect
@@ -255,7 +257,6 @@ require (
 	go.opentelemetry.io/otel/metric v1.38.0 // indirect
 	go.opentelemetry.io/otel/trace v1.38.0 // indirect
 	go.shabbyrobe.org/gocovmerge v0.0.0-20230507111327-fa4f82cfbf4d // indirect
-	go.uber.org/dig v1.19.0 // indirect
 	go.uber.org/multierr v1.11.0 // indirect
 	go.uber.org/zap v1.27.0 // indirect
 	go.yaml.in/yaml/v2 v2.4.3 // indirect
@@ -263,7 +264,6 @@ require (
 	golang.org/x/exp v0.0.0-20251023183803-a4bb9ffd2546 // indirect
 	golang.org/x/net v0.47.0 // indirect
 	golang.org/x/oauth2 v0.33.0 // indirect
-	golang.org/x/sys v0.38.0 // indirect
 	golang.org/x/text v0.31.0 // indirect
 	golang.org/x/time v0.14.0 // indirect
 	golang.org/x/tools v0.38.0 // indirect

internal/blob/packer.go

@@ -1,14 +1,16 @@
 // Package blob handles the creation of blobs - the final storage units for Vaultik.
 // A blob is a large file (up to 10GB) containing many compressed and encrypted chunks
-// from multiple source files. Blobs are content-addressed, meaning their filename
-// is derived from the SHA256 hash of their compressed and encrypted content.
+// from multiple source files. Blobs are content-addressed: a blob's filename is
+// hex(SHA256(SHA256(uncompressed blob contents))), computed from the concatenated
+// chunk data before compression and encryption, not from the stored bytes. See
+// blobgen.DoubleSHA256 and docs/REPOSTRUCTURE.md.
 //
 // The blob creation process:
 //  1. Chunks are accumulated from multiple files
-// 2. The collection is compressed using zstd
-// 3. The compressed data is encrypted using age
-// 4. The encrypted blob is hashed to create its content-addressed name
-// 5. The blob is uploaded to S3 using the hash as the filename
+//  2. Each chunk's uncompressed bytes are fed to a running SHA-256 and, in the same
+//     pass, compressed with zstd and encrypted with age into the temp file
+//  3. On finalize, the name is the double SHA-256 of that uncompressed content
+//  4. The blob is uploaded to S3 using the name as the filename
 //
 // This design optimizes storage efficiency by batching many small chunks into
 // larger blobs, reducing the number of S3 operations and associated costs.
@@ -243,6 +245,17 @@ func (p *Packer) Flush(ctx context.Context) error {
 	return nil
 }
 
+// DiscardCurrentBlob closes and removes the temp file of the blob being
+// packed, if any, and drops that blob. A caller that stops before Flush
+// must call it, or the temp file is left behind. Thread-safe.
+func (p *Packer) DiscardCurrentBlob() {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+
+	p.cleanupTempFile()
+	p.currentBlob = nil
+}
+
 // FinalizeBlob finalizes the current blob being assembled.
 // This compresses the accumulated chunks, encrypts the result, and computes
 // the content-addressed hash. The finalized blob is either passed to the
@@ -487,7 +500,7 @@ func (p *Packer) closeBlobWriter() (string, int64, error) {
 		return "", 0, fmt.Errorf("seeking to start: %w", err)
 	}
 
-	finalHash := p.currentBlob.writer.Sum256()
+	finalHash := p.currentBlob.writer.ContentID()
 
 	return hex.EncodeToString(finalHash), finalSize, nil
 }

internal/blobgen/compress.go

@@ -1,89 +0,0 @@
-// Package blobgen implements the blob data pipeline: streaming zstd
-// compression, age encryption, and SHA256 content hashing for blob
-// creation, plus the matching decrypt/decompress/verify reader.
-package blobgen
-
-import (
-	"bytes"
-	"encoding/hex"
-	"fmt"
-	"io"
-)
-
-// CompressResult contains the results of compression
-type CompressResult struct {
-	Data             []byte
-	UncompressedSize int64
-	CompressedSize   int64
-	SHA256           string
-}
-
-// CompressData compresses and encrypts data, returning the result with hash
-func CompressData(
-	data []byte, compressionLevel int, recipients []string,
-) (*CompressResult, error) {
-	var buf bytes.Buffer
-
-	// Create writer
-	w, err := NewWriter(&buf, compressionLevel, recipients)
-	if err != nil {
-		return nil, fmt.Errorf("creating writer: %w", err)
-	}
-
-	// Write data
-	_, err = w.Write(data)
-	if err != nil {
-		_ = w.Close()
-
-		return nil, fmt.Errorf("writing data: %w", err)
-	}
-
-	// Close to flush
-	err = w.Close()
-	if err != nil {
-		return nil, fmt.Errorf("closing writer: %w", err)
-	}
-
-	return &CompressResult{
-		Data:             buf.Bytes(),
-		UncompressedSize: int64(len(data)),
-		CompressedSize:   int64(buf.Len()),
-		SHA256:           hex.EncodeToString(w.Sum256()),
-	}, nil
-}
-
-// CompressStream compresses and encrypts from reader to writer, returning
-// the number of uncompressed bytes written and the content hash.
-func CompressStream(
-	dst io.Writer, src io.Reader, compressionLevel int, recipients []string,
-) (int64, string, error) {
-	// Create writer
-	w, err := NewWriter(dst, compressionLevel, recipients)
-	if err != nil {
-		return 0, "", fmt.Errorf("creating writer: %w", err)
-	}
-
-	closed := false
-
-	defer func() {
-		if !closed {
-			_ = w.Close()
-		}
-	}()
-
-	// Copy data
-	_, err = io.Copy(w, src)
-	if err != nil {
-		return 0, "", fmt.Errorf("copying data: %w", err)
-	}
-
-	// Close to flush
-	err = w.Close()
-	if err != nil {
-		return 0, "", fmt.Errorf("closing writer: %w", err)
-	}
-
-	closed = true
-
-	return w.BytesWritten(), hex.EncodeToString(w.Sum256()), nil
-}

internal/blobgen/compress_test.go

@@ -1,80 +0,0 @@
-package blobgen_test
-
-import (
-	"bytes"
-	"crypto/rand"
-	"strings"
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-	"sneak.berlin/go/vaultik/internal/blobgen"
-)
-
-// testRecipient is a static age recipient for tests.
-const testRecipient = "age1cplgrwj77ta54dnmydvvmzn64ltk83ankxl5sww04mrtmu62kv3s89gmvv"
-
-// TestCompressStreamNoDoubleClose is a regression test for issue #28.
-// It verifies that CompressStream does not panic or return an error due to
-// double-closing the underlying blobgen.Writer. Before the fix in PR #33,
-// the explicit Close() on the happy path combined with defer Close() would
-// cause a double close.
-func TestCompressStreamNoDoubleClose(t *testing.T) {
-	t.Parallel()
-
-	input := []byte("regression test data for issue #28 double-close fix")
-
-	var buf bytes.Buffer
-
-	written, hash, err := blobgen.CompressStream(
-		&buf, bytes.NewReader(input), 3, []string{testRecipient})
-	require.NoError(t, err, "CompressStream should not return an error")
-	assert.Positive(t, written, "expected bytes written > 0")
-	assert.NotEmpty(t, hash, "expected non-empty hash")
-	assert.Positive(t, buf.Len(), "expected non-empty output")
-}
-
-// TestCompressStreamLargeInput exercises CompressStream with a larger payload
-// to ensure no double-close issues surface under heavier I/O.
-func TestCompressStreamLargeInput(t *testing.T) {
-	t.Parallel()
-
-	data := make([]byte, 512*1024) // 512 KB
-	_, err := rand.Read(data)
-	require.NoError(t, err)
-
-	var buf bytes.Buffer
-
-	written, hash, err := blobgen.CompressStream(
-		&buf, bytes.NewReader(data), 3, []string{testRecipient})
-	require.NoError(t, err)
-	assert.Positive(t, written)
-	assert.NotEmpty(t, hash)
-}
-
-// TestCompressStreamEmptyInput verifies CompressStream handles empty input
-// without double-close issues.
-func TestCompressStreamEmptyInput(t *testing.T) {
-	t.Parallel()
-
-	var buf bytes.Buffer
-
-	_, hash, err := blobgen.CompressStream(
-		&buf, strings.NewReader(""), 3, []string{testRecipient})
-	require.NoError(t, err)
-	assert.NotEmpty(t, hash)
-}
-
-// TestCompressDataNoDoubleClose mirrors the stream test for CompressData,
-// ensuring the explicit Close + error-path Close pattern is also safe.
-func TestCompressDataNoDoubleClose(t *testing.T) {
-	t.Parallel()
-
-	input := []byte("CompressData regression test for double-close")
-
-	result, err := blobgen.CompressData(input, 3, []string{testRecipient})
-	require.NoError(t, err)
-	assert.Positive(t, result.CompressedSize)
-	assert.Equal(t, result.UncompressedSize, int64(len(input)))
-	assert.NotEmpty(t, result.SHA256)
-}

internal/blobgen/helpers_test.go

@@ -0,0 +1,119 @@
+package blobgen_test
+
+import (
+	"bytes"
+	"crypto/rand"
+	"errors"
+	"io"
+	"testing"
+
+	"filippo.io/age"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/blobgen"
+)
+
+// ageChunkSize is age's STREAM plaintext chunk size (64 KiB); each encrypted
+// chunk adds a 16-byte ChaCha20-Poly1305 tag.
+const (
+	ageChunkSize    = 64 * 1024
+	ageChunkTagSize = 16
+	ageSegmentSize  = ageChunkSize + ageChunkTagSize
+	ageNonceSize    = 16
+)
+
+// makeIdentity returns a fresh X25519 identity and its recipient string.
+func makeIdentity(t *testing.T) (*age.X25519Identity, string) {
+	t.Helper()
+
+	id, err := age.GenerateX25519Identity()
+	require.NoError(t, err)
+
+	return id, id.Recipient().String()
+}
+
+// randomBytes returns n cryptographically random bytes, which do not compress
+// so the encrypted payload spans multiple age segments.
+func randomBytes(t *testing.T, n int) []byte {
+	t.Helper()
+
+	b := make([]byte, n)
+	_, err := rand.Read(b)
+	require.NoError(t, err)
+
+	return b
+}
+
+// compressibleBytes returns n bytes of a repeating pattern, which zstd packs
+// down to a small payload.
+func compressibleBytes(n int) []byte {
+	pattern := bytes.Repeat([]byte("compressible-"), n/13+1)
+
+	return pattern[:n]
+}
+
+// encryptBlob compresses, encrypts and returns a blob for plaintext at
+// compression level 1.
+func encryptBlob(t *testing.T, plaintext []byte, recipients ...string) []byte {
+	t.Helper()
+
+	var buf bytes.Buffer
+
+	w, err := blobgen.NewWriter(&buf, 1, recipients)
+	require.NoError(t, err)
+
+	_, err = w.Write(plaintext)
+	require.NoError(t, err)
+	require.NoError(t, w.Close())
+
+	return buf.Bytes()
+}
+
+// ageHeaderLen returns the byte length of blob's age header, i.e. the offset
+// of the 16-byte payload nonce that follows it. The header ends with a MAC
+// line "--- <mac>\n"; the nonce begins right after that newline.
+func ageHeaderLen(t *testing.T, blob []byte) int {
+	t.Helper()
+
+	i := bytes.Index(blob, []byte("\n--- "))
+	require.GreaterOrEqual(t, i, 0, "age MAC footer line not found")
+
+	nl := bytes.IndexByte(blob[i+1:], '\n')
+	require.GreaterOrEqual(t, nl, 0, "newline ending MAC line not found")
+
+	return i + 1 + nl + 1
+}
+
+// requireBlobUnreadable asserts that data never decrypts to a plaintext with a
+// nil error: either NewReader fails, or reading it does.
+func requireBlobUnreadable(t *testing.T, data []byte, id age.Identity) {
+	t.Helper()
+
+	r, err := blobgen.NewReader(bytes.NewReader(data), id)
+	if err != nil {
+		return
+	}
+
+	_, err = io.ReadAll(r)
+	_ = r.Close()
+
+	require.Error(t, err, "reading a damaged blob must fail")
+}
+
+// errFailWriter is returned by failAfterWriter once its byte limit is passed.
+var errFailWriter = errors.New("destination write failed")
+
+// failAfterWriter accepts writes until more than limit bytes have been sent,
+// then fails every write. It models a destination that dies mid-blob.
+type failAfterWriter struct {
+	limit   int
+	written int
+}
+
+func (f *failAfterWriter) Write(p []byte) (int, error) {
+	f.written += len(p)
+	if f.written > f.limit {
+		return 0, errFailWriter
+	}
+
+	return len(p), nil
+}

internal/blobgen/limit.go

@@ -0,0 +1,49 @@
+package blobgen
+
+import (
+	"errors"
+	"io"
+)
+
+// ErrOutputTooLarge is returned by a reader from LimitReader once it has
+// been asked for more than its limit. It bounds how far an untrusted
+// compressed stream may expand, so a small, highly compressible object
+// from the store cannot decompress without limit.
+var ErrOutputTooLarge = errors.New("output exceeds size limit")
+
+// LimitReader returns a reader that yields at most limit bytes from r and
+// then fails with ErrOutputTooLarge. Unlike io.LimitReader, which reports
+// a silent io.EOF at the limit (indistinguishable from a stream that
+// simply ended), this fails, so a caller decoding or copying the stream
+// sees an error rather than a truncated value. A stream of exactly limit
+// bytes reads back cleanly to EOF; the first byte beyond it is the error.
+func LimitReader(r io.Reader, limit int64) io.Reader {
+	// remaining counts down from limit+1: the extra byte is the one that,
+	// if it ever arrives, proves the stream is longer than the limit.
+	return &limitReader{r: r, remaining: limit + 1}
+}
+
+type limitReader struct {
+	r         io.Reader
+	remaining int64
+}
+
+func (l *limitReader) Read(p []byte) (int, error) {
+	if l.remaining <= 0 {
+		return 0, ErrOutputTooLarge
+	}
+
+	if int64(len(p)) > l.remaining {
+		p = p[:l.remaining]
+	}
+
+	n, err := l.r.Read(p)
+	l.remaining -= int64(n)
+
+	if l.remaining <= 0 {
+		// The (limit+1)th byte was just read: the stream is too long.
+		return n, ErrOutputTooLarge
+	}
+
+	return n, err
+}

internal/blobgen/limit_test.go

@@ -0,0 +1,43 @@
+package blobgen_test
+
+import (
+	"bytes"
+	"io"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/blobgen"
+)
+
+// TestLimitReaderPassesExactSize checks that a stream of exactly the limit
+// reads back cleanly to EOF: the bound must not reject a legitimate blob
+// whose plaintext equals its recorded size.
+func TestLimitReaderPassesExactSize(t *testing.T) {
+	t.Parallel()
+
+	const n = 1000
+
+	r := blobgen.LimitReader(bytes.NewReader(bytes.Repeat([]byte("a"), n)), n)
+
+	got, err := io.ReadAll(r)
+	require.NoError(t, err)
+	require.Len(t, got, n)
+}
+
+// TestLimitReaderFailsPastLimit feeds a large, highly compressible run of
+// zeros — the decompressed output a zip bomb would produce — through a
+// small limit and checks it fails within the bound rather than passing
+// the whole stream through.
+func TestLimitReaderFailsPastLimit(t *testing.T) {
+	t.Parallel()
+
+	const limit = 1000
+
+	r := blobgen.LimitReader(
+		bytes.NewReader(bytes.Repeat([]byte{0}, limit*1000)), limit)
+
+	n, err := io.Copy(io.Discard, r)
+	require.ErrorIs(t, err, blobgen.ErrOutputTooLarge)
+	require.LessOrEqual(t, n, int64(limit)+1,
+		"reader must stop within one byte of the limit")
+}

internal/blobgen/negative_test.go

@@ -0,0 +1,190 @@
+package blobgen_test
+
+import (
+	"bytes"
+	"fmt"
+	"testing"
+
+	"filippo.io/age"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/blobgen"
+)
+
+// TestNewReaderWrongIdentity covers issue case 4: opening a blob with an
+// identity other than the recipient reports no matching identity.
+func TestNewReaderWrongIdentity(t *testing.T) {
+	t.Parallel()
+
+	_, recipient := makeIdentity(t)
+	other, _ := makeIdentity(t)
+
+	blob := encryptBlob(t, []byte("secret payload"), recipient)
+
+	_, err := blobgen.NewReader(bytes.NewReader(blob), other)
+	require.Error(t, err)
+
+	var noMatch *age.NoIdentityMatchError
+	assert.ErrorAs(t, err, &noMatch)
+}
+
+// TestNewReaderTruncated covers issue case 6: a multi-segment blob cut at
+// several points must never read back as valid data. The point immediately
+// after the header and nonce is intentionally excluded: it reads as a valid
+// empty blob today and is the regression case for
+// https://git.eeqj.de/sneak/vaultik/issues/152.
+func TestNewReaderTruncated(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+	blob := encryptBlob(t, randomBytes(t, 4*65536+123), recipient)
+	h := ageHeaderLen(t, blob)
+
+	require.Greater(t, len(blob), h+ageNonceSize+ageSegmentSize,
+		"test needs a blob of at least two age segments")
+
+	cases := []struct {
+		name string
+		size int
+	}{
+		{"inside header", h / 2},
+		{"inside nonce", h + 8},
+		{"inside first segment", h + ageNonceSize + 100},
+		{"end of first full segment", h + ageNonceSize + ageSegmentSize},
+		{"last byte removed", len(blob) - 1},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+			requireBlobUnreadable(t, blob[:tc.size], id)
+		})
+	}
+}
+
+// TestNewReaderCorrupted covers issue case 7: one flipped byte in each region
+// of a multi-segment blob makes it unreadable.
+func TestNewReaderCorrupted(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+	blob := encryptBlob(t, randomBytes(t, 4*65536+123), recipient)
+	h := ageHeaderLen(t, blob)
+
+	firstNL := bytes.IndexByte(blob, '\n')
+	require.Positive(t, firstNL, "header must have a version line")
+
+	cases := []struct {
+		name string
+		pos  int
+	}{
+		{"header stanza", firstNL + 5},
+		{"header MAC line", h - 2},
+		{"nonce", h + 4},
+		{"body segment", h + ageNonceSize + 50},
+		{"final tag", len(blob) - 1},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			corrupt := append([]byte(nil), blob...)
+			corrupt[tc.pos] ^= 0xff
+			requireBlobUnreadable(t, corrupt, id)
+		})
+	}
+}
+
+// TestNewReaderTrailingAndGarbage covers issue case 8: bytes appended after a
+// valid blob, empty input, and random garbage each fail to read.
+func TestNewReaderTrailingAndGarbage(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+
+	valid := encryptBlob(t, []byte("small payload"), recipient)
+	appended := append(append([]byte(nil), valid...), []byte("trailing junk")...)
+
+	t.Run("appended bytes", func(t *testing.T) {
+		t.Parallel()
+		requireBlobUnreadable(t, appended, id)
+	})
+	t.Run("empty input", func(t *testing.T) {
+		t.Parallel()
+		requireBlobUnreadable(t, []byte{}, id)
+	})
+	t.Run("random garbage", func(t *testing.T) {
+		t.Parallel()
+		requireBlobUnreadable(t, randomBytes(t, 512), id)
+	})
+}
+
+// TestNewWriterInvalidLevel covers the rejected end of issue case 9: an
+// out-of-range compression level errors and writes nothing to the destination.
+func TestNewWriterInvalidLevel(t *testing.T) {
+	t.Parallel()
+
+	_, recipient := makeIdentity(t)
+
+	for _, level := range []int{0, -1, 20} {
+		t.Run(fmt.Sprintf("level%d", level), func(t *testing.T) {
+			t.Parallel()
+
+			var buf bytes.Buffer
+
+			w, err := blobgen.NewWriter(&buf, level, []string{recipient})
+			require.ErrorIs(t, err, blobgen.ErrInvalidCompressionLevel)
+			assert.Nil(t, w)
+			assert.Zero(t, buf.Len(), "nothing written on an invalid level")
+		})
+	}
+}
+
+// TestNewWriterInvalidRecipients covers issue case 10: nil and empty recipient
+// lists and an unparsable recipient string each error.
+func TestNewWriterInvalidRecipients(t *testing.T) {
+	t.Parallel()
+
+	cases := []struct {
+		name       string
+		recipients []string
+	}{
+		{"nil list", nil},
+		{"empty list", []string{}},
+		{"invalid recipient string", []string{"not-a-recipient"}},
+	}
+
+	for _, tc := range cases {
+		t.Run(tc.name, func(t *testing.T) {
+			t.Parallel()
+
+			var buf bytes.Buffer
+
+			w, err := blobgen.NewWriter(&buf, 1, tc.recipients)
+			require.Error(t, err)
+			assert.Nil(t, w)
+		})
+	}
+}
+
+// TestNewWriterFailingDestination covers issue case 11: a destination that
+// fails mid-blob surfaces its error from Write or Close.
+func TestNewWriterFailingDestination(t *testing.T) {
+	t.Parallel()
+
+	_, recipient := makeIdentity(t)
+
+	// The limit clears the age header and nonce so NewWriter succeeds, then
+	// trips once the compressed body starts flowing.
+	dst := &failAfterWriter{limit: 512}
+
+	w, err := blobgen.NewWriter(dst, 1, []string{recipient})
+	require.NoError(t, err)
+
+	_, writeErr := w.Write(randomBytes(t, 256*1024))
+	closeErr := w.Close()
+
+	assert.True(t, writeErr != nil || closeErr != nil,
+		"destination failure must surface from Write or Close")
+}

internal/blobgen/reader.go

@@ -2,6 +2,7 @@ package blobgen
 
 import (
 	"crypto/sha256"
+	"errors"
 	"fmt"
 	"hash"
 	"io"
@@ -20,10 +21,12 @@ type Reader struct {
 	bytesRead    int64
 }
 
-// NewReader creates a new Reader that decrypts, decompresses, and verifies data
-func NewReader(r io.Reader, identity age.Identity) (*Reader, error) {
+// NewReader creates a new Reader that decrypts, decompresses, and verifies
+// data. Every supplied identity is offered to age.Decrypt, so a blob
+// encrypted to any one of them can be read.
+func NewReader(r io.Reader, identities ...age.Identity) (*Reader, error) {
 	// Create decryption reader
-	decReader, err := age.Decrypt(r, identity)
+	decReader, err := age.Decrypt(r, identities...)
 	if err != nil {
 		return nil, fmt.Errorf("creating decryption reader: %w", err)
 	}
@@ -54,6 +57,22 @@ func (r *Reader) Read(p []byte) (int, error) {
 	n, err := r.teeReader.Read(p)
 	r.bytesRead += int64(n)
 
+	// When the ciphertext is cut right after the age header plus its
+	// 16-byte nonce, the age reader's first read fails with
+	// io.ErrUnexpectedEOF, and the zstd decoder maps that to a clean
+	// io.EOF at frame start. That makes a truncated stream look like a
+	// valid empty one. Distinguish the two: on EOF, read once more from
+	// the age reader. A genuine end leaves it at (0, io.EOF); a truncated
+	// stream leaves its stored io.ErrUnexpectedEOF, which we surface.
+	if errors.Is(err, io.EOF) {
+		var probe [1]byte
+
+		m, ageErr := r.decryptor.Read(probe[:])
+		if m != 0 || !errors.Is(ageErr, io.EOF) {
+			return n, io.ErrUnexpectedEOF
+		}
+	}
+
 	return n, err
 }
 
@@ -64,7 +83,9 @@ func (r *Reader) Close() error {
 	return nil
 }
 
-// Sum256 returns the SHA256 hash of all data read
+// Sum256 returns the single SHA-256 of the plaintext read so far. This is the
+// first hash only; the stored object name is its double hash, which callers
+// obtain by passing this digest to DoubleSHA256.
 func (r *Reader) Sum256() []byte {
 	return r.hasher.Sum(nil)
 }

internal/blobgen/reader_test.go

@@ -0,0 +1,54 @@
+package blobgen_test
+
+import (
+	"bytes"
+	"io"
+	"testing"
+
+	"filippo.io/age"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/blobgen"
+)
+
+// TestMultipleRecipients verifies that data written for several recipients can
+// be read back by each recipient's identity. Moved from internal/crypto, which
+// held the only multi-recipient test; blobgen is now the sole encryption path.
+func TestMultipleRecipients(t *testing.T) {
+	t.Parallel()
+
+	identities := make([]*age.X25519Identity, 3)
+	recipients := make([]string, 3)
+
+	for i := range identities {
+		identity, err := age.GenerateX25519Identity()
+		require.NoError(t, err)
+
+		identities[i] = identity
+		recipients[i] = identity.Recipient().String()
+	}
+
+	plaintext := []byte("Secret message for multiple recipients")
+
+	var encrypted bytes.Buffer
+
+	writer, err := blobgen.NewWriter(&encrypted, 3, recipients)
+	require.NoError(t, err)
+	_, err = writer.Write(plaintext)
+	require.NoError(t, err)
+	require.NoError(t, writer.Close())
+
+	// Every recipient's identity must recover the original plaintext.
+	for i, identity := range identities {
+		reader, err := blobgen.NewReader(
+			bytes.NewReader(encrypted.Bytes()), identity)
+		require.NoError(t, err, "recipient %d should open the reader", i+1)
+
+		got, err := io.ReadAll(reader)
+		require.NoError(t, err, "recipient %d should read the plaintext", i+1)
+		require.NoError(t, reader.Close())
+
+		assert.Equal(t, plaintext, got,
+			"recipient %d should recover the original plaintext", i+1)
+	}
+}

internal/blobgen/roundtrip_test.go

@@ -0,0 +1,132 @@
+package blobgen_test
+
+import (
+	"bytes"
+	"crypto/sha256"
+	"fmt"
+	"io"
+	"testing"
+
+	"filippo.io/age"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/blobgen"
+)
+
+// checkRoundTrip writes input through a Writer, reads it back through a Reader,
+// and verifies the plaintext, the byte counts, and the content hashes.
+func checkRoundTrip(
+	t *testing.T, id *age.X25519Identity, recipient string,
+	level int, input []byte,
+) {
+	t.Helper()
+
+	var buf bytes.Buffer
+
+	w, err := blobgen.NewWriter(&buf, level, []string{recipient})
+	require.NoError(t, err)
+
+	n, err := w.Write(input)
+	require.NoError(t, err)
+	assert.Equal(t, len(input), n)
+	require.NoError(t, w.Close())
+	require.Equal(t, int64(len(input)), w.BytesWritten())
+
+	r, err := blobgen.NewReader(bytes.NewReader(buf.Bytes()), id)
+	require.NoError(t, err)
+
+	got, err := io.ReadAll(r)
+	require.NoError(t, err)
+	require.NoError(t, r.Close())
+
+	assert.Equal(t, input, got, "decrypted output must equal input")
+	require.Equal(t, int64(len(input)), r.BytesRead())
+
+	// The hash values are checked by decrypting: the reader's single SHA-256
+	// is the hash of the plaintext, and hashing it once more (DoubleSHA256)
+	// gives the writer's ContentID.
+	single := sha256.Sum256(got)
+	assert.Equal(t, single[:], r.Sum256())
+	assert.Equal(t, blobgen.DoubleSHA256(r.Sum256()), w.ContentID())
+}
+
+// TestWriterReaderRoundTrip covers issue cases 1 and 2: every size round trips
+// for both random and compressible data, and the reader hash, its double hash
+// and the byte counts all agree.
+func TestWriterReaderRoundTrip(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+
+	// Sizes exercise the age segment boundary (64 KiB) from just below to a
+	// few segments above it, plus the empty and single-byte edges.
+	sizes := []int{0, 1, 65535, 65536, 65537, 4*65536 + 123}
+
+	kinds := []struct {
+		name string
+		fill func(*testing.T, int) []byte
+	}{
+		{"random", randomBytes},
+		{"compressible", func(_ *testing.T, n int) []byte {
+			return compressibleBytes(n)
+		}},
+	}
+
+	for _, k := range kinds {
+		for _, size := range sizes {
+			name := fmt.Sprintf("%s/%d", k.name, size)
+			t.Run(name, func(t *testing.T) {
+				t.Parallel()
+				checkRoundTrip(t, id, recipient, 1, k.fill(t, size))
+			})
+		}
+	}
+}
+
+// TestZeroLengthNoWrite covers issue case 3: a Writer closed with no Write at
+// all produces the double hash of the empty input, and the blob reads back as
+// empty with no error.
+func TestZeroLengthNoWrite(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+
+	var buf bytes.Buffer
+
+	w, err := blobgen.NewWriter(&buf, 1, []string{recipient})
+	require.NoError(t, err)
+	require.NoError(t, w.Close())
+	assert.Equal(t, int64(0), w.BytesWritten())
+
+	empty := sha256.Sum256(nil)
+	doubled := sha256.Sum256(empty[:])
+	assert.Equal(t, doubled[:], w.ContentID(),
+		"ContentID of empty input is SHA256(SHA256(\"\"))")
+
+	r, err := blobgen.NewReader(bytes.NewReader(buf.Bytes()), id)
+	require.NoError(t, err)
+
+	got, err := io.ReadAll(r)
+	require.NoError(t, err)
+	require.NoError(t, r.Close())
+
+	assert.Empty(t, got, "empty blob decrypts to empty output")
+	assert.Equal(t, int64(0), r.BytesRead())
+	assert.Equal(t, empty[:], r.Sum256())
+}
+
+// TestNewWriterValidLevelsRoundTrip covers the accepted end of issue case 9:
+// the boundary compression levels 1 and 19 both round trip.
+func TestNewWriterValidLevelsRoundTrip(t *testing.T) {
+	t.Parallel()
+
+	id, recipient := makeIdentity(t)
+	input := randomBytes(t, 4096)
+
+	for _, level := range []int{1, 19} {
+		t.Run(fmt.Sprintf("level%d", level), func(t *testing.T) {
+			t.Parallel()
+			checkRoundTrip(t, id, recipient, level, input)
+		})
+	}
+}

internal/blobgen/truncation_test.go

@@ -0,0 +1,70 @@
+package blobgen_test
+
+import (
+	"bytes"
+	"io"
+	"testing"
+
+	"filippo.io/age"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/blobgen"
+)
+
+// TestReaderRejectsHeaderNonceTruncation guards against a stream cut right
+// after the age header plus its 16-byte nonce. age.Decrypt still succeeds on
+// such an object, and the zstd decoder maps the age reader's
+// io.ErrUnexpectedEOF to a clean io.EOF at frame start, so without the extra
+// check the truncated stream would read as a valid empty one. Reading it must
+// now fail.
+func TestReaderRejectsHeaderNonceTruncation(t *testing.T) {
+	t.Parallel()
+
+	identity, err := age.GenerateX25519Identity()
+	require.NoError(t, err)
+
+	// Encrypting empty plaintext yields header + nonce(16) + a single
+	// 16-byte final chunk tag. Dropping the trailing tag leaves exactly the
+	// age header plus its nonce — the truncation point that triggers the bug.
+	var full bytes.Buffer
+
+	w, err := age.Encrypt(&full, identity.Recipient())
+	require.NoError(t, err)
+	require.NoError(t, w.Close())
+
+	truncated := full.Bytes()[:full.Len()-16]
+
+	reader, err := blobgen.NewReader(bytes.NewReader(truncated), identity)
+	require.NoError(t, err)
+
+	defer func() { _ = reader.Close() }()
+
+	_, err = io.ReadAll(reader)
+	require.Error(t, err)
+	require.ErrorIs(t, err, io.ErrUnexpectedEOF)
+}
+
+// TestReaderReadsGenuinelyEmptyBlob confirms the truncation check does not
+// reject a legitimately empty payload: a blob written with no data must round
+// trip back to zero bytes with no error.
+func TestReaderReadsGenuinelyEmptyBlob(t *testing.T) {
+	t.Parallel()
+
+	identity, err := age.GenerateX25519Identity()
+	require.NoError(t, err)
+
+	var encrypted bytes.Buffer
+
+	writer, err := blobgen.NewWriter(
+		&encrypted, 3, []string{identity.Recipient().String()})
+	require.NoError(t, err)
+	require.NoError(t, writer.Close())
+
+	reader, err := blobgen.NewReader(bytes.NewReader(encrypted.Bytes()), identity)
+	require.NoError(t, err)
+
+	defer func() { _ = reader.Close() }()
+
+	data, err := io.ReadAll(reader)
+	require.NoError(t, err)
+	require.Empty(t, data)
+}

internal/blobgen/writer.go

@@ -1,3 +1,6 @@
+// Package blobgen implements the blob data pipeline: streaming zstd
+// compression, age encryption, and SHA256 content hashing for blob
+// creation, plus the matching decrypt/decompress/verify reader.
 package blobgen
 
 import (
@@ -12,6 +15,24 @@ import (
 	"github.com/klauspost/compress/zstd"
 )
 
+// DoubleSHA256 returns the double SHA-256 of content whose single SHA-256
+// digest is sum: it hashes that digest once more. Stored objects — a blob, and
+// the metadata database export — are named by this second hash.
+//
+// The second hash does not hide whether known content is stored: an attacker
+// who can reproduce an object's entire plaintext computes the same name simply
+// by hashing twice, exactly as this code does. What limits that is blob
+// packing, not the double hash — a blob's name covers all of its concatenated
+// chunk plaintext, so a name can be confirmed only by someone who can
+// reproduce the whole blob (a snapshot made entirely of known content, or a
+// known file large enough to fill blobs on its own). An ordinary file that
+// shares a blob with other, unknown data cannot be confirmed this way.
+func DoubleSHA256(sum []byte) []byte {
+	h := sha256.Sum256(sum)
+
+	return h[:]
+}
+
 // Zstd compression level bounds accepted by NewWriter.
 const (
 	minCompressionLevel = 1
@@ -27,6 +48,11 @@ const reservedCompressionCPUs = 2
 var ErrInvalidCompressionLevel = errors.New(
 	"invalid compression level: must be between 1 and 19")
 
+// errInvalidRecipient is returned when a recipient string does not parse as
+// an X25519 age1... public key. It omits the value, which can be sensitive.
+var errInvalidRecipient = errors.New(
+	"not a valid X25519 age1... recipient")
+
 // Writer wraps compression and encryption with SHA256 hashing.
 // Data flows: input -> tee(hasher, compressor -> encryptor -> destination)
 // The hash is computed on the uncompressed input for deterministic content-addressing.
@@ -57,10 +83,12 @@ func NewWriter(
 	// Parse recipients
 	var ageRecipients []age.Recipient
 
-	for _, recipient := range recipients {
+	for i, recipient := range recipients {
+		// The recipient string can be sensitive (e.g. a secret key pasted by
+		// mistake), so the error names its position, never its value.
 		r, err := age.ParseX25519Recipient(recipient)
 		if err != nil {
-			return nil, fmt.Errorf("parsing recipient %s: %w", recipient, err)
+			return nil, fmt.Errorf("%w: recipient %d", errInvalidRecipient, i)
 		}
 
 		ageRecipients = append(ageRecipients, r)
@@ -123,17 +151,12 @@ func (w *Writer) Close() error {
 	return nil
 }
 
-// Sum256 returns the double SHA256 hash of the uncompressed input data.
-// Double hashing (SHA256(SHA256(data))) prevents information leakage about
-// the plaintext - an attacker cannot confirm existence of known content
-// by computing its hash and checking for a matching blob filename.
-func (w *Writer) Sum256() []byte {
-	// First hash: SHA256(plaintext)
-	firstHash := w.hasher.Sum(nil)
-	// Second hash: SHA256(firstHash) - this is the blob ID
-	secondHash := sha256.Sum256(firstHash)
-
-	return secondHash[:]
+// ContentID returns the double SHA-256 of the uncompressed input data: the
+// name under which this content is stored. It is the second hash of the
+// running SHA-256, via DoubleSHA256; see that function for what naming content
+// this way does and does not hide.
+func (w *Writer) ContentID() []byte {
+	return DoubleSHA256(w.hasher.Sum(nil))
 }
 
 // BytesWritten returns the number of uncompressed bytes written

internal/blobgen/writer_test.go

@@ -12,9 +12,10 @@ import (
 	"sneak.berlin/go/vaultik/internal/blobgen"
 )
 
-// TestWriterHashIsDoubleHash verifies that Writer.Sum256() returns
-// the double hash SHA256(SHA256(plaintext)) for security.
-// Double hashing prevents attackers from confirming existence of known content.
+// TestWriterHashIsDoubleHash verifies that Writer.ContentID() returns
+// SHA256(SHA256(plaintext)). Stored objects are named by this second hash so a
+// name is not the plaintext's own SHA-256; this does not stop someone who
+// already holds the plaintext from confirming it.
 func TestWriterHashIsDoubleHash(t *testing.T) {
 	t.Parallel()
 
@@ -43,7 +44,7 @@ func TestWriterHashIsDoubleHash(t *testing.T) {
 	require.NoError(t, err)
 
 	// Get the hash from the writer
-	writerHash := hex.EncodeToString(writer.Sum256())
+	writerHash := hex.EncodeToString(writer.ContentID())
 
 	// Calculate the expected double hash: SHA256(SHA256(plaintext))
 	firstHash := sha256.Sum256(testData)
@@ -60,11 +61,11 @@ func TestWriterHashIsDoubleHash(t *testing.T) {
 
 	// The writer hash should match the double hash
 	assert.Equal(t, expectedDoubleHash, writerHash,
-		"Writer.Sum256() should return SHA256(SHA256(plaintext)) for security")
+		"Writer.ContentID() must be SHA256(SHA256(plaintext))")
 
-	// Verify it's NOT the single hash (would leak information)
+	// It must be the second hash, not the plaintext's own SHA-256.
 	assert.NotEqual(t, singleHashStr, writerHash,
-		"Writer hash should not be single hash (would allow content confirmation attacks)")
+		"Writer hash must be the double hash, not the single SHA-256")
 }
 
 // TestWriterDeterministicHash verifies that the same input always produces
@@ -93,8 +94,8 @@ func TestWriterDeterministicHash(t *testing.T) {
 	require.NoError(t, err)
 	require.NoError(t, writer2.Close())
 
-	hash1 := hex.EncodeToString(writer1.Sum256())
-	hash2 := hex.EncodeToString(writer2.Sum256())
+	hash1 := hex.EncodeToString(writer1.ContentID())
+	hash2 := hex.EncodeToString(writer2.ContentID())
 
 	// Hashes should be identical (deterministic)
 	assert.Equal(t, hash1, hash2, "Same input should produce same hash")
@@ -108,3 +109,20 @@ func TestWriterDeterministicHash(t *testing.T) {
 	t.Logf("Encrypted size 1: %d bytes", buf1.Len())
 	t.Logf("Encrypted size 2: %d bytes", buf2.Len())
 }
+
+// TestNewWriterSecretKeyNotEchoed verifies that a secret key mistakenly passed
+// as a recipient does not appear in the returned error. A recipient string can
+// be sensitive, so the error must name only the position, not the value.
+func TestNewWriterSecretKeyNotEchoed(t *testing.T) {
+	t.Parallel()
+
+	secretKey := "AGE-SECRET-KEY-19CR5YSFW59HM4TLD6GX" +
+		"VEDMZFTVVF7PPHKUT68TXSFPK7APHXA2QS2NJA5"
+
+	var buf bytes.Buffer
+
+	_, err := blobgen.NewWriter(&buf, 3, []string{secretKey})
+	require.Error(t, err)
+	assert.NotContains(t, err.Error(), secretKey,
+		"error must not echo the recipient value")
+}

internal/chunker/chunker.go

@@ -33,9 +33,10 @@ type Chunker struct {
 	maxChunkSize int
 }
 
-// chunkSizeSpread is the FastCDC-recommended factor between the average
+// ChunkSizeSpread is the FastCDC-recommended factor between the average
 // chunk size and the minimum (avg/spread) and maximum (avg*spread) sizes.
-const chunkSizeSpread = 4
+// The largest chunk the chunker can emit is therefore avg*ChunkSizeSpread.
+const ChunkSizeSpread = 4
 
 // NewChunker creates a new chunker with the specified average chunk size.
 // The actual chunk sizes will vary between avgChunkSize/4 and avgChunkSize*4
@@ -45,8 +46,8 @@ func NewChunker(avgChunkSize int64) *Chunker {
 	// FastCDC recommends min = avg/4 and max = avg*4
 	return &Chunker{
 		avgChunkSize: int(avgChunkSize),
-		minChunkSize: int(avgChunkSize / chunkSizeSpread),
-		maxChunkSize: int(avgChunkSize * chunkSizeSpread),
+		minChunkSize: int(avgChunkSize / ChunkSizeSpread),
+		maxChunkSize: int(avgChunkSize * ChunkSizeSpread),
 	}
 }
 

internal/cli/app.go

@@ -7,15 +7,13 @@ import (
 	"context"
 	"errors"
 	"fmt"
-	"os"
-	"os/signal"
 	"path/filepath"
-	"strings"
-	"syscall"
+	"sync"
 	"time"
 
 	"github.com/adrg/xdg"
 	"github.com/spf13/cobra"
+	"go.uber.org/dig"
 	"go.uber.org/fx"
 	"sneak.berlin/go/vaultik/internal/config"
 	"sneak.berlin/go/vaultik/internal/database"
@@ -32,14 +30,33 @@ import (
 // may take before we give up.
 const shutdownTimeout = 30 * time.Second
 
-// AppOptions contains common options for creating the fx application.
-// It includes the configuration file path, logging options, and additional
-// fx modules and invocations that should be included in the application.
+// lockMode says whether a command mutates persistent state — the local
+// index database or the remote store — and so must hold the process-wide
+// PID lock, or only reads that state and may run alongside a mutator.
+type lockMode int
+
+const (
+	// mutating commands (snapshot create, snapshot purge, snapshot remove,
+	// prune, remote nuke) write the local index or the remote store. They
+	// hold the PID lock so that at most one runs at a time.
+	mutating lockMode = iota
+	// readOnly commands (info, snapshot list, snapshot verify, remote info,
+	// snapshot restore) do not write the local index or the remote store,
+	// so they run without the lock and are never blocked by a running
+	// mutator. restore writes only to the target directory it is given.
+	readOnly
+)
+
+// AppOptions contains common options for creating and running the fx
+// application: the configuration file path, logging options, additional fx
+// modules and invocations, and whether the command mutates persistent
+// state (which decides whether it takes the PID lock).
 type AppOptions struct {
 	ConfigPath string
 	LogOptions log.Options
 	Modules    []fx.Option
 	Invokes    []fx.Option
+	Mode       lockMode
 }
 
 // setupGlobals records the startup time and, when an output-suppression
@@ -48,6 +65,11 @@ type AppOptions struct {
 // silenced — per the documented convention that --quiet suppresses
 // non-error output only. The startup banner is printed by Entry
 // before cobra parses arguments, gated by the same arg-level check.
+//
+// --json quiets the UI here too, because stdout then carries a JSON
+// document and human narration would corrupt it. Unlike Quiet it does
+// not lower the stderr log level (issue #112), so --verbose/--debug
+// still surface diagnostics alongside the document.
 func setupGlobals(
 	lc fx.Lifecycle, g *globals.Globals, v *vaultik.Vaultik, opts log.Options,
 ) {
@@ -55,7 +77,7 @@ func setupGlobals(
 		OnStart: func(_ context.Context) error {
 			g.StartTime = time.Now().UTC()
 
-			if opts.Cron || opts.Quiet {
+			if opts.Cron || opts.Quiet || opts.JSON {
 				v.UI.SetQuiet(true)
 			}
 
@@ -124,87 +146,180 @@ func (e *startupError) Error() string {
 //	could not build arguments for function "Y" (file:line): failed to build U:
 //	received non-nil error from function "Z" (file:line): <real error>
 //
-// Users care about the real error, not the DI plumbing. We strip everything
-// up through the last "): " (which is always the close-paren of an fx
-// function-location annotation followed by the wrapped error).
+// Users care about the real error, not the DI plumbing. dig.RootCause,
+// from the library fx is built on, unwraps those layers and returns the
+// real error whole. Cutting the text at "): " instead would also cut a
+// real error that contains it, such as the blob_size_limit one.
 func cleanStartupError(err error) error {
-	msg := err.Error()
-	if idx := strings.LastIndex(msg, "): "); idx >= 0 {
-		msg = msg[idx+3:]
-	}
-
-	return &startupError{msg: msg}
+	return &startupError{msg: dig.RootCause(err).Error()}
 }
 
-// RunApp starts and stops the fx application within the given context.
-// It handles graceful shutdown on interrupt signals (SIGINT, SIGTERM) and
-// ensures the application stops cleanly. The function blocks until the
-// application completes or is interrupted. Returns an error if startup fails.
+// RunApp starts the fx application, blocks until it is asked to stop, and
+// then stops it. The app is asked to stop either by an OS interrupt
+// (SIGINT/SIGTERM — fx installs its own handler when app.Wait is called) or,
+// on normal completion, by the finished operation calling
+// Shutdowner.Shutdown(); both arrive on the app.Wait channel.
+//
+// Stopping runs the fx OnStop hooks, and RunApp does not return until Stop
+// returns. On an interrupt the operation's OnStop hook cancels the running
+// command and waits for it to unwind — removing its decrypted scratch files —
+// so the process cannot proceed to exit mid-cleanup (issue #159). Waiting for
+// Stop before returning is what makes that hook effective: routing the
+// interrupt through app.Stop and not returning until it completes is required,
+// because fx also fires the app.Wait channel on the signal, and an earlier
+// version returned on that alone — unwinding to os.Exit while the concurrent
+// cleanup still ran. The stop is bounded by shutdownTimeout. Returns an error
+// if startup fails.
 func RunApp(ctx context.Context, app *fx.App) error {
-	// Set up signal handling for graceful shutdown
-	sigChan := make(chan os.Signal, 1)
-	signal.Notify(sigChan, os.Interrupt, syscall.SIGTERM)
-
-	// Create a context that will be cancelled on signal
-	ctx, cancel := context.WithCancel(ctx)
-	defer cancel()
-
-	// Start the app
 	err := app.Start(ctx)
 	if err != nil {
 		return cleanStartupError(err)
 	}
 
-	// Handle shutdown
-	shutdownComplete := make(chan struct{})
+	// Block until an interrupt or the finished operation's
+	// Shutdowner.Shutdown() arrives, then stop the app in this goroutine so we
+	// return only after its OnStop hooks — including the operation's cleanup
+	// wait — have run. Detach the stop from ctx's cancellation but keep its
+	// values, and bound it by shutdownTimeout.
+	<-app.Wait()
 
-	go func() {
-		defer close(shutdownComplete)
-
-		<-sigChan
-		log.Notice("Received interrupt signal, shutting down gracefully...")
-
-		// Create a timeout context for shutdown. The parent ctx is being
-		// cancelled, so detach from its cancellation but keep its values.
-		shutdownCtx, shutdownCancel := context.WithTimeout(
+	shutdownCtx, cancel := context.WithTimeout(
 		context.WithoutCancel(ctx), shutdownTimeout)
-		defer shutdownCancel()
+	defer cancel()
 
-		err := app.Stop(shutdownCtx)
+	err = app.Stop(shutdownCtx)
 	if err != nil {
 		log.Error("Error during shutdown", "error", err)
 	}
-	}()
 
-	// Wait for the signal handler to complete shutdown or the app to
-	// request shutdown.
-	select {
-	case <-shutdownComplete:
-		// Shutdown completed via signal
 	return nil
-	case <-ctx.Done():
-		// Context cancelled (shouldn't happen in normal operation)
-		err := app.Stop(context.WithoutCancel(ctx))
+}
+
+// errReported marks a failure the operation has already shown the user
+// (or, under `snapshot verify --json`, put in its document). Entry
+// turns it into a non-zero exit status without printing anything
+// further, so the error line is not doubled. It flows up from
+// RunOperation through cobra to Entry.
+var errReported = errors.New("operation failed")
+
+// errInterrupted marks an operation that SIGINT or SIGTERM stopped
+// before it finished. Entry shows it and returns exitCodeInterrupted.
+var errInterrupted = errors.New("interrupted before the command finished")
+
+// RunOperation runs op against the Vaultik instance inside the fx app
+// and turns a failure into a returned error rather than an os.Exit from
+// within the goroutine. An os.Exit there skipped main's deferred
+// profile writers -- so profiling a failing command yielded a truncated
+// profile (issue #75) -- and RunWithApp's PID-lock release, and denied
+// the app any graceful shutdown; returning the error to the top runs
+// all three.
+//
+// op runs in a goroutine so OnStart returns promptly and an interrupt
+// can still cancel through OnStop; when it finishes, success or failure,
+// it triggers shutdown, which is what lets RunWithApp return. On an
+// interrupt OnStop cancels op and waits for the goroutine to return, so
+// op's cleanup (removing decrypted scratch files) runs before the
+// process exits; the wait is bounded by shutdownTimeout. report is
+// called with a failure so the caller can show it to the user before
+// it becomes errReported.
+//
+// The run counts as interrupted unless op returned, without an
+// interrupt having cancelled it, before RunWithApp returned. An
+// interrupted op is not reported, whatever it returned; RunOperation
+// returns errInterrupted instead.
+func RunOperation(
+	ctx context.Context, opts AppOptions,
+	op func(v *vaultik.Vaultik) error, report func(err error),
+) error {
+	var (
+		mu       sync.Mutex
+		finished bool // op returned before any interrupt cancelled it
+		failed   bool // op finished with an error
+	)
+
+	opts.Invokes = append(opts.Invokes,
+		fx.Invoke(func(v *vaultik.Vaultik, lc fx.Lifecycle) {
+			var stop func(context.Context) bool
+
+			lc.Append(fx.Hook{
+				OnStart: func(_ context.Context) error {
+					stop = v.StartOperation(func() {
+						err := op(v)
+
+						// Only stop, called from OnStop below, cancels the
+						// Vaultik context, so a live context means no
+						// interrupt cancelled op. Check the context, not
+						// err: an interrupted op need not return
+						// context.Canceled (`snapshot verify --json`
+						// returns a verification failure).
+						if v.Context().Err() == nil {
 							if err != nil {
-			log.Error("Error stopping app", "error", err)
+								report(err)
 							}
 
-		return ctx.Err()
-	case <-app.Done():
-		// App finished running (e.g., backup completed)
+							mu.Lock()
+							finished = true
+							failed = err != nil
+							mu.Unlock()
+						}
+
+						stopErr := v.Shutdowner.Shutdown()
+						if stopErr != nil {
+							log.Error("Failed to shutdown", "error", stopErr)
+						}
+					})
+
+					return nil
+				},
+				// On an interrupt, cancel the operation and wait for it to
+				// unwind so its cleanup defers (which remove decrypted
+				// scratch files from the temp directory) run before the
+				// process exits. The wait is bounded by ctx, the existing
+				// shutdownTimeout.
+				OnStop: func(ctx context.Context) error {
+					if !stop(ctx) {
+						log.Warn("Shutdown timed out before the operation " +
+							"finished; decrypted temporary files may remain")
+					}
+
+					return nil
+				},
+			})
+		}))
+
+	err := RunWithApp(ctx, opts)
+	if err != nil {
+		return err
+	}
+
+	// RunWithApp returns only after the app was asked to stop, either by
+	// an interrupt or by the goroutine's Shutdown call. When op finished
+	// without being cancelled, the goroutine set finished before that
+	// call. So if finished is unset here, an interrupt stopped the app,
+	// and op either returned after it was cancelled or is still running
+	// because the shutdown timed out.
+	mu.Lock()
+	defer mu.Unlock()
+
+	switch {
+	case !finished:
+		return errInterrupted
+	case failed:
+		return errReported
+	default:
 		return nil
 	}
 }
 
 // runVaultikApp runs the standard single-operation command lifecycle
-// shared by the list/purge/verify/remove/remote-info subcommands:
-// resolve the config, start the fx app, run op against the Vaultik
-// instance in a goroutine, report a failure prefixed with failMsg
-// (suppressed while suppressErrors is true, e.g. under --json), then
-// trigger shutdown. The operation is cancelled when the app stops.
-// extraQuiet is OR-ed into LogOptions.Quiet (e.g. --json output modes).
+// shared by the snapshot list/purge/remove and remote nuke subcommands:
+// resolve the config, then run op against the Vaultik instance through
+// RunOperation, reporting a failure prefixed with failMsg on stderr. mode
+// says whether the command takes the PID lock. jsonOutput marks a command
+// whose stdout is a JSON document: it quiets the UI but, unlike Quiet,
+// leaves the stderr log level alone.
 func runVaultikApp(
-	cmd *cobra.Command, extraQuiet, suppressErrors bool,
+	cmd *cobra.Command, mode lockMode, jsonOutput bool,
 	failMsg string, op func(v *vaultik.Vaultik) error,
 ) error {
 	configPath, err := ResolveConfigPath()
@@ -214,75 +329,64 @@ func runVaultikApp(
 
 	rootFlags := GetRootFlags()
 
-	return RunWithApp(cmd.Context(), AppOptions{
+	return RunOperation(cmd.Context(), AppOptions{
 		ConfigPath: configPath,
 		LogOptions: log.Options{
 			Verbose: rootFlags.Verbose,
 			Debug:   rootFlags.Debug,
-			Quiet:   rootFlags.Quiet || extraQuiet,
+			Quiet:   rootFlags.Quiet,
+			JSON:    jsonOutput,
 		},
-		Modules: []fx.Option{},
-		Invokes: []fx.Option{
-			fx.Invoke(func(v *vaultik.Vaultik, lc fx.Lifecycle) {
-				lc.Append(fx.Hook{
-					OnStart: func(_ context.Context) error {
-						go func() {
-							err := op(v)
-							if err != nil {
-								if !errors.Is(err, context.Canceled) {
-									if !suppressErrors {
+		Mode: mode,
+	}, op, func(err error) {
 		log.Error(failMsg, "error", err)
 		ReportErrorf("%s: %v", failMsg, err)
-									}
-
-									os.Exit(1)
-								}
-							}
-
-							err = v.Shutdowner.Shutdown()
-							if err != nil {
-								log.Error("Failed to shutdown", "error", err)
-							}
-						}()
-
-						return nil
-					},
-					OnStop: func(_ context.Context) error {
-						v.Cancel()
-
-						return nil
-					},
-				})
-			}),
-		},
 	})
 }
 
 // RunWithApp is a helper that creates and runs an fx app with the given options.
 // It combines NewApp and RunApp into a single convenient function. This is the
 // preferred way to run CLI commands that need the full application context.
-// It acquires a PID lock before starting to prevent concurrent instances.
+// A mutating command takes the process-wide PID lock before starting so that
+// only one runs at a time; a read-only command runs without it and is not
+// blocked while a mutator holds the lock (opts.Mode).
 func RunWithApp(ctx context.Context, opts AppOptions) error {
-	// Acquire PID lock to prevent concurrent instances
-	lockDir := filepath.Join(xdg.DataHome, "vaultik")
-
-	lock, err := pidlock.Acquire(lockDir)
+	release, err := acquireLockIfMutating(opts.Mode,
+		filepath.Join(xdg.DataHome, "vaultik"))
 	if err != nil {
-		if errors.Is(err, pidlock.ErrAlreadyRunning) {
-			return fmt.Errorf("cannot start: %w", err)
+		return err
 	}
 
-		return fmt.Errorf("failed to acquire lock: %w", err)
-	}
-
-	defer func() {
-		err := lock.Release()
-		if err != nil {
-			log.Warn("Failed to release PID lock", "error", err)
-		}
-	}()
+	defer release()
 
 	app := NewApp(opts)
 
 	return RunApp(ctx, app)
 }
+
+// acquireLockIfMutating takes the process-wide PID lock in lockDir for a
+// mutating command and returns a function that releases it. A read-only
+// command takes no lock, so it returns a no-op release and is never blocked
+// while a mutator holds the lock. ErrAlreadyRunning (another mutator holds
+// the lock) is surfaced as a "cannot start" error.
+func acquireLockIfMutating(mode lockMode, lockDir string) (func(), error) {
+	if mode != mutating {
+		return func() {}, nil
+	}
+
+	lock, err := pidlock.Acquire(lockDir)
+	if err != nil {
+		if errors.Is(err, pidlock.ErrAlreadyRunning) {
+			return nil, fmt.Errorf("cannot start: %w", err)
+		}
+
+		return nil, fmt.Errorf("failed to acquire lock: %w", err)
+	}
+
+	return func() {
+		err := lock.Release()
+		if err != nil {
+			log.Warn("Failed to release PID lock", "error", err)
+		}
+	}, nil
+}

internal/cli/app_test.go

@@ -2,54 +2,103 @@ package cli //nolint:testpackage // needs access to unexported cleanStartupError
 
 import (
 	"errors"
+	"os"
+	"path/filepath"
 	"testing"
+
+	"go.uber.org/fx"
+	"sneak.berlin/go/vaultik/internal/config"
+	"sneak.berlin/go/vaultik/internal/pidlock"
 )
 
+// TestCleanStartupError starts a real fx app whose config has a
+// blob_size_limit below the minimum. fx puts annotations naming the
+// functions it was building in front of the config error; only those
+// may be removed, although the config error itself contains "): ".
 func TestCleanStartupError(t *testing.T) {
 	t.Parallel()
 
-	tests := []struct {
-		name string
-		in   string
-		want string
-	}{
-		{
-			name: "real fx error chain",
-			in: `could not build arguments for function ` +
-				`"sneak.berlin/go/vaultik/internal/cli".newSnapshotCreateCommand.func1.1 ` +
-				`(/Users/user/dev/vaultik/internal/cli/snapshot.go:71): ` +
-				`failed to build *vaultik.Vaultik: ` +
-				`could not build arguments for function ` +
-				`"sneak.berlin/go/vaultik/internal/vaultik".New ` +
-				`(/Users/user/dev/vaultik/internal/vaultik/vaultik.go:59): ` +
-				`failed to build storage.Storer: ` +
-				`received non-nil error from function ` +
-				`"sneak.berlin/go/vaultik/internal/storage".NewStorer ` +
-				`(/Users/user/dev/vaultik/internal/storage/module.go:23): ` +
-				`creating base path: mkdir /Volumes/BACKUPS: permission denied`,
-			want: `creating base path: mkdir /Volumes/BACKUPS: permission denied`,
-		},
-		{
-			name: "no fx wrapping",
-			in:   "plain error",
-			want: "plain error",
-		},
-		{
-			name: "single fx wrapping",
-			in:   `received non-nil error from function "foo" (file.go:1): underlying problem`,
-			want: "underlying problem",
-		},
+	path := filepath.Join(t.TempDir(), "config.yml")
+
+	err := os.WriteFile(path, []byte("snapshots:\n"+
+		"  test:\n"+
+		"    paths: [/tmp/vaultik-test-source]\n"+
+		"storage_url: file:///tmp/vaultik-test-storage\n"+
+		"chunk_size: 10MB\n"+
+		"blob_size_limit: 30MB\n"), 0o600)
+	if err != nil {
+		t.Fatalf("write config: %v", err)
 	}
 
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			t.Parallel()
+	app := fx.New(
+		fx.NopLogger,
+		fx.Supply(config.Path(path)),
+		config.Module,
+		fx.Invoke(func(*config.Config) {}),
+	)
 
-			//nolint:err113 // test constructs errors from table input
-			got := cleanStartupError(errors.New(tt.in)).Error()
-			if got != tt.want {
-				t.Errorf("got %q, want %q", got, tt.want)
+	err = app.Start(t.Context())
+	if err == nil {
+		t.Fatal("app started; want a blob_size_limit error")
 	}
-		})
+
+	want := "failed to load config: invalid config: blob_size_limit must " +
+		"be at least the largest chunk the chunker can emit (chunk_size " +
+		"times the FastCDC size spread): need at least 40000000 bytes"
+
+	got := cleanStartupError(err).Error()
+	if got != want {
+		t.Errorf("got %q, want %q", got, want)
 	}
 }
+
+// TestCleanStartupErrorNotFromFx checks that an error fx did not wrap is
+// shown unchanged.
+func TestCleanStartupErrorNotFromFx(t *testing.T) {
+	t.Parallel()
+
+	//nolint:err113 // a one-off error standing in for a startup failure
+	got := cleanStartupError(errors.New("plain error")).Error()
+	if got != "plain error" {
+		t.Errorf("got %q, want %q", got, "plain error")
+	}
+}
+
+// TestLockScopedToMutatingCommands proves the partition the PID lock now
+// enforces: a read-only command runs while a mutator holds the lock, and
+// two mutating commands still mutually exclude.
+func TestLockScopedToMutatingCommands(t *testing.T) {
+	t.Parallel()
+
+	lockDir := filepath.Join(t.TempDir(), "vaultik")
+
+	// A mutating command takes the process-wide lock.
+	releaseMutator, err := acquireLockIfMutating(mutating, lockDir)
+	if err != nil {
+		t.Fatalf("mutating command could not acquire lock: %v", err)
+	}
+
+	// A read-only command runs to completion even while the lock is held.
+	releaseReader, err := acquireLockIfMutating(readOnly, lockDir)
+	if err != nil {
+		t.Fatalf("read-only command was blocked by held lock: %v", err)
+	}
+
+	releaseReader()
+
+	// A second mutating command is refused while the first holds the lock.
+	_, err = acquireLockIfMutating(mutating, lockDir)
+	if !errors.Is(err, pidlock.ErrAlreadyRunning) {
+		t.Fatalf("second mutating command was not excluded, got: %v", err)
+	}
+
+	// Once the first mutator releases, another mutating command may run.
+	releaseMutator()
+
+	release, err := acquireLockIfMutating(mutating, lockDir)
+	if err != nil {
+		t.Fatalf("mutating command could not acquire released lock: %v", err)
+	}
+
+	release()
+}

internal/cli/config.go

@@ -1,16 +1,22 @@
 package cli
 
 import (
+	"bytes"
 	"errors"
 	"fmt"
 	"os"
 	"os/exec"
 	"path/filepath"
+	"reflect"
+	"slices"
 	"strconv"
 	"strings"
+	"unicode/utf8"
 
 	"github.com/spf13/cobra"
 	"gopkg.in/yaml.v3"
+	"sneak.berlin/go/vaultik/internal/config"
+	"sneak.berlin/go/vaultik/internal/ui"
 )
 
 // configFileMode is the permission set for freshly written config files;
@@ -24,6 +30,14 @@ const configSetArgs = 2
 // parent config dirs (e.g. ~/.config) are conventionally traversable.
 const configDirMode = 0o755
 
+// configYAMLIndent matches the 2-space indentation of defaultConfigTemplate,
+// so `config set` writes the file back with the same indentation rather than
+// yaml.Marshal's 4-space default.
+const configYAMLIndent = 2
+
+// yamlStringTag is YAML's tag for a string scalar.
+const yamlStringTag = "!!str"
+
 var (
 	errConfigExists     = errors.New("config file already exists")
 	errEmptyConfig      = errors.New("empty config file")
@@ -31,6 +45,9 @@ var (
 	errNeedNumericIndex = errors.New("key is a list; use a numeric index")
 	errIndexOutOfRange  = errors.New("index out of range")
 	errNotMapOrList     = errors.New("key is not a map or list")
+	errWouldNotLoad     = errors.New(
+		"vaultik would not load the config after this change, " +
+			"so the file was not changed")
 )
 
 const defaultConfigTemplate = `# vaultik configuration
@@ -38,13 +55,19 @@ const defaultConfigTemplate = `# vaultik configuration
 
 # ─── REQUIRED ────────────────────────────────────────────────────────────────
 
-# Age recipient public keys for encryption.
-# Backups are encrypted to ALL listed recipients. Any one of the corresponding
-# private keys can decrypt. Generate a keypair with:
+# Age recipient public keys for encryption. snapshot create needs at least
+# one; listing, verifying and restoring do not, so a machine that only
+# restores can leave this empty.
+# Backups are encrypted to ALL listed recipients; any one of the corresponding
+# private keys can decrypt. Adding a recipient later does not re-encrypt data
+# already stored: deduplicated chunks and existing blobs stay encrypted to the
+# earlier recipients, so a newly added key cannot restore them on its own (see
+# docs/REPOSTRUCTURE.md, Accepted Risks). Generate a keypair and add its
+# public key with:
 #   age-keygen -o vaultik_backup_private_key.txt
 #   grep 'public key' vaultik_backup_private_key.txt
-age_recipients:
-  - age1REPLACE_WITH_YOUR_PUBLIC_KEY
+#   vaultik config set age_recipients.0 age1...
+age_recipients: []
 
 # Named snapshots. Each snapshot backs up one or more paths and can have its
 # own exclude patterns in addition to the global excludes below.
@@ -186,8 +209,8 @@ storage_url: ""
 #   access_key_id: YOUR_ACCESS_KEY
 #   secret_access_key: YOUR_SECRET_KEY
 #   # region: us-east-1        # Default: us-east-1
-#   # use_ssl: true            # Default: true
-#   # part_size: 5MB           # Multipart upload part size. Default: 5MB
+#   # part_size: 5MiB          # Upload part size, 5MiB to 5GiB. Default: 5MiB
+#   # For the s3:// form, disable TLS with ?ssl=false in the URL, not use_ssl.
 
 # ─── OPTIONAL ────────────────────────────────────────────────────────────────
 
@@ -201,11 +224,13 @@ storage_url: ""
 
 # Average chunk size for content-defined chunking (FastCDC).
 # Smaller = better deduplication but more metadata overhead.
-# Accepts: 1MB, 10M, 64KB, etc.
+# Minimum 1MB. Accepts: 1MB, 10M, 64MiB, etc.
 # Default: 10MB
 # chunk_size: 10MB
 
 # Maximum blob size before splitting into a new blob.
+# Must be at least four times chunk_size (the largest chunk the chunker can
+# emit); a smaller limit would let a single-chunk blob exceed it.
 # Accepts: 1GB, 10G, 500MB, etc.
 # Default: 10GB
 # blob_size_limit: 10GB
@@ -214,6 +239,18 @@ storage_url: ""
 # Default: 3
 # compression_level: 3
 
+# How old a blob that no snapshot lists must be before prune deletes it,
+# and how old every object in a directory under metadata/ with no manifest
+# must be before prune deletes that directory. A backup uploads its blobs
+# and its db.zst.age before its manifest, so on a destination shared
+# between hosts or users this must be longer than the longest backup run.
+# It does not protect a blob a backup reuses from an earlier run that
+# stopped before its manifest: that blob keeps the time of its first
+# upload, so a prune can delete it.
+# Accepts: 36h, 90m, etc. 0s deletes such blobs and directories at once.
+# Default: 24h
+# prune_grace_period: 24h
+
 # Hostname used in snapshot IDs. Default: system hostname.
 # hostname: myserver
 
@@ -230,6 +267,12 @@ func NewConfigCommand() *cobra.Command {
 		Use:   "config",
 		Short: "Manage the configuration file",
 		Long:  "Commands for creating, editing, and querying the vaultik config file.",
+		// cobra checks Args only on a command that has a Run: without
+		// one, a mistyped subcommand printed help and exited 0.
+		Args: cobra.NoArgs,
+		Run: func(cmd *cobra.Command, _ []string) {
+			_ = cmd.Help()
+		},
 	}
 
 	cmd.AddCommand(newConfigInitCommand())
@@ -246,14 +289,16 @@ func newConfigInitCommand() *cobra.Command {
 		Use:   "init",
 		Short: "Write a default config file",
 		Long: `Creates a default configuration file with commented explanations
-for every setting. If a config file already exists at the target path,
-the command refuses to overwrite it.
+for the settings it lists. If a config file already exists at the target
+path, the command refuses to overwrite it. The file is created with mode
+0600, because it can hold credentials.
 
 The config is written to the path from --config, $VAULTIK_CONFIG, or
-the platform default config directory (e.g. ~/Library/Application Support/
-on macOS, ~/.config/ on Linux, /etc/vaultik/ as root).`,
+config.yml in the platform default config directory
+(~/Library/Application Support/vaultik/ on macOS, ~/.config/vaultik/ on
+Linux, /etc/vaultik/ as root).`,
 		Args: cobra.NoArgs,
-		RunE: func(_ *cobra.Command, _ []string) error {
+		RunE: func(cmd *cobra.Command, _ []string) error {
 			path := configPathForInit()
 
 			_, err := os.Stat(path)
@@ -273,8 +318,11 @@ on macOS, ~/.config/ on Linux, /etc/vaultik/ as root).`,
 				return fmt.Errorf("writing config file: %w", err)
 			}
 
-			_, _ = fmt.Fprintf(os.Stdout, "Config written to %s\n", path)
-			_, _ = fmt.Fprintln(os.Stdout,
+			// A written-confirmation, not scriptable output: route it
+			// through the UI so it is styled and --quiet silences it.
+			out := commandUI(cmd)
+			out.Infof("Config written to %s.", path)
+			out.Infof(
 				"Edit it to set your age_recipients, snapshots, and storage_url.")
 
 			return nil
@@ -286,7 +334,7 @@ on macOS, ~/.config/ on Linux, /etc/vaultik/ as root).`,
 func newConfigEditCommand() *cobra.Command {
 	return &cobra.Command{
 		Use:   "edit",
-		Short: "Open the config file in $EDITOR",
+		Short: "Open the config file in $EDITOR (vi when it is not set)",
 		Args:  cobra.NoArgs,
 		RunE: func(cmd *cobra.Command, _ []string) error {
 			path, err := ResolveConfigPath()
@@ -299,8 +347,15 @@ func newConfigEditCommand() *cobra.Command {
 				editor = "vi"
 			}
 
+			// Run $EDITOR through the shell, as git does, so a value
+			// with arguments such as "code --wait" works. The path
+			// goes in as an argument ("$@"), not into the command
+			// text, so a path with spaces or quotes arrives whole.
+			// sh -c takes the word after the command as $0, which
+			// is why editor appears twice.
 			//nolint:gosec // G204: launching the operator's own $EDITOR is the point
-			ed := exec.CommandContext(cmd.Context(), editor, path)
+			ed := exec.CommandContext(cmd.Context(),
+				"sh", "-c", editor+` "$@"`, editor, path)
 			ed.Stdin = os.Stdin
 			ed.Stdout = os.Stdout
 			ed.Stderr = os.Stderr
@@ -316,7 +371,7 @@ func newConfigGetCommand() *cobra.Command {
 		Use:   "get <key>",
 		Short: "Print a config value by dotted path (e.g. storage_url, compression_level)",
 		Args:  cobra.ExactArgs(1),
-		RunE: func(_ *cobra.Command, args []string) error {
+		RunE: func(cmd *cobra.Command, args []string) error {
 			path, err := ResolveConfigPath()
 			if err != nil {
 				return err
@@ -332,8 +387,13 @@ func newConfigGetCommand() *cobra.Command {
 				return err
 			}
 
+			// The value is scriptable output: it must stay machine-plain
+			// (no marker, no color) and is never silenced by --quiet, so it
+			// is written straight to stdout rather than through the UI.
+			w := cmd.OutOrStdout()
+
 			if node.Kind == yaml.ScalarNode {
-				_, _ = fmt.Fprintln(os.Stdout, node.Value)
+				_, _ = fmt.Fprintln(w, node.Value)
 
 				return nil
 			}
@@ -343,7 +403,7 @@ func newConfigGetCommand() *cobra.Command {
 				return fmt.Errorf("marshaling value: %w", err)
 			}
 
-			_, _ = fmt.Fprint(os.Stdout, string(out))
+			_, _ = fmt.Fprint(w, string(out))
 
 			return nil
 		},
@@ -356,8 +416,17 @@ func newConfigSetCommand() *cobra.Command {
 		Use:   "set <key> <value>",
 		Short: "Set a config value by dotted path (e.g. compression_level 5)",
 		Long: `Sets a scalar config value addressed by dotted YAML path and writes
-the file back, preserving comments and formatting. Intermediate maps
-are created as needed.
+the file back, preserving comments and formatting. Missing keys along
+the path are created: as a list when the next path element is a
+number, otherwise as a map.
+
+If the config loads, a change after which it would not, such as
+compression_level 99, is refused and the file is left unchanged. A
+config that does not load, such as the one config init writes before
+storage_url is set or one with a ${ENV:...} reference to a variable
+not set here, is checked one setting at a time: only a change after
+which the changed setting cannot be read, such as one giving it the
+wrong type, is refused.
 
 Examples:
   vaultik config set storage_url "file:///mnt/backups"
@@ -365,44 +434,154 @@ Examples:
   vaultik config set compression_level 9
   vaultik config set s3.bucket mybucket  # legacy S3 fields still supported`,
 		Args: cobra.ExactArgs(configSetArgs),
-		RunE: func(_ *cobra.Command, args []string) error {
+		RunE: func(cmd *cobra.Command, args []string) error {
 			path, err := ResolveConfigPath()
 			if err != nil {
 				return err
 			}
 
+			return writeConfigSet(commandUI(cmd), path, args[0], args[1])
+		},
+	}
+}
+
+// writeConfigSet applies key=value to the config at path, writes it back
+// owner-only, and confirms the write by naming just the key through the
+// UI writer (styled, and silenced by --quiet). The value is never
+// echoed: it may be a secret such as s3.secret_access_key, and captured
+// stdout or a pasted terminal would then leak it.
+func writeConfigSet(out *ui.Writer, path, key, value string) error {
 	root, err := loadYAMLFile(path)
 	if err != nil {
 		return err
 	}
 
-			err = yamlPathSet(root, strings.Split(args[0], "."), args[1])
-			if err != nil {
-				return err
-			}
-
-			out, err := yaml.Marshal(root)
+	oldData, err := marshalConfigYAML(root)
 	if err != nil {
 		return fmt.Errorf("marshaling config: %w", err)
 	}
 
-			mode := os.FileMode(configFileMode)
+	keys := strings.Split(key, ".")
 
-			info, statErr := os.Stat(path)
-			if statErr == nil {
-				mode = info.Mode().Perm()
+	err = yamlPathSet(root, keys, value)
+	if err != nil {
+		return err
 	}
 
-			err = os.WriteFile(path, out, mode)
+	data, err := marshalConfigYAML(root)
+	if err != nil {
+		return fmt.Errorf("marshaling config: %w", err)
+	}
+
+	setting, err := yamlPathOnly(root, keys)
+	if err != nil {
+		return err
+	}
+
+	settingData, err := marshalConfigYAML(setting)
+	if err != nil {
+		return fmt.Errorf("marshaling config: %w", err)
+	}
+
+	err = checkConfigChange(oldData, data, settingData)
+	if err != nil {
+		return fmt.Errorf("%w: %w", errWouldNotLoad, err)
+	}
+
+	err = os.WriteFile(path, data, configFileMode)
 	if err != nil {
 		return fmt.Errorf("writing config file: %w", err)
 	}
 
-			_, _ = fmt.Fprintf(os.Stdout, "%s = %s\n", args[0], args[1])
+	// os.WriteFile does not change the mode of a file that already exists,
+	// so a config that was group- or world-readable stays that way. As it
+	// may hold S3 credentials, tighten it to owner-only after writing.
+	info, statErr := os.Stat(path)
+	if statErr == nil && info.Mode().Perm()&0o044 != 0 {
+		err = os.Chmod(path, configFileMode)
+		if err != nil {
+			return fmt.Errorf("tightening config file permissions: %w", err)
+		}
+	}
+
+	out.Infof("Set %s.", key)
 
 	return nil
-		},
+}
+
+// checkConfigChange returns the reason vaultik would not load newData, the
+// config after a change, if the change brings that failure in. If oldData,
+// the config before it, loads, the new one must load too. A config that
+// does not load, such as one still being filled in or one with a ${...}
+// reference that cannot be filled in here, is checked one setting at a
+// time: settingData, the changed setting on its own, must decode into the
+// config type.
+func checkConfigChange(oldData, newData, settingData []byte) error {
+	oldCfg, err := config.Parse(oldData)
+	if err == nil && oldCfg.Validate() == nil {
+		newCfg, err := config.Parse(newData)
+		if err != nil {
+			return err
 		}
+
+		return newCfg.Validate()
+	}
+
+	_, err = config.Parse(settingData)
+
+	return err
+}
+
+// yamlPathOnly returns a document holding only the value at the dotted key
+// path in root, inside the maps and lists that lead to it there.
+func yamlPathOnly(root *yaml.Node, keys []string) (*yaml.Node, error) {
+	node, err := yamlPathGet(root, keys)
+	if err != nil {
+		return nil, err
+	}
+
+	for i, key := range slices.Backward(keys) {
+		parent, err := yamlPathGet(root, keys[:i])
+		if err != nil {
+			return nil, err
+		}
+
+		if parent.Kind == yaml.SequenceNode {
+			node = &yaml.Node{
+				Kind: yaml.SequenceNode, Content: []*yaml.Node{node},
+			}
+		} else {
+			keyNode := &yaml.Node{Kind: yaml.ScalarNode, Value: key}
+			node = &yaml.Node{
+				Kind: yaml.MappingNode, Content: []*yaml.Node{keyNode, node},
+			}
+		}
+	}
+
+	return &yaml.Node{Kind: yaml.DocumentNode, Content: []*yaml.Node{node}}, nil
+}
+
+// marshalConfigYAML renders a config document tree with 2-space indentation,
+// matching defaultConfigTemplate. yaml.Marshal defaults to 4 spaces, which
+// would reindent the whole file on the first `config set` despite the promise
+// to preserve formatting.
+func marshalConfigYAML(root *yaml.Node) ([]byte, error) {
+	var buf bytes.Buffer
+
+	enc := yaml.NewEncoder(&buf)
+	enc.SetIndent(configYAMLIndent)
+
+	err := enc.Encode(root)
+	if err != nil {
+		return nil, err
+	}
+
+	err = enc.Close()
+	if err != nil {
+		return nil, err
+	}
+
+	return buf.Bytes(), nil
 }
 
 // loadYAMLFile parses a YAML file into a yaml.Node document tree,
@@ -488,8 +667,8 @@ func yamlPathGet(root *yaml.Node, keys []string) (*yaml.Node, error) {
 	return node, nil
 }
 
-// yamlPathSet navigates a dotted key path, creating intermediate maps as
-// needed, and sets the final key to the given scalar value. Numeric path
+// yamlPathSet navigates a dotted key path, creating missing keys as needed,
+// and sets the final key to the given scalar value. Numeric path
 // components index into sequences; an index equal to the sequence length
 // appends a new element (e.g. "age_recipients.1" on a 1-element list).
 func yamlPathSet(root *yaml.Node, keys []string, value string) error {
@@ -502,12 +681,12 @@ func yamlPathSet(root *yaml.Node, keys []string, value string) error {
 		node = node.Content[0]
 	}
 
-	for i, key := range keys {
+	for i := range keys {
 		last := i == len(keys)-1
 
 		switch node.Kind {
 		case yaml.MappingNode:
-			node = yamlSetInMapping(node, key, value, last)
+			node = yamlSetInMapping(node, keys, i, value, last)
 		case yaml.SequenceNode:
 			next, err := yamlSetInSequence(node, keys, i, value, last)
 			if err != nil {
@@ -524,17 +703,68 @@ func yamlPathSet(root *yaml.Node, keys []string, value string) error {
 		}
 	}
 
+	// config.Load reads the file through untyped YAML, which turns an
+	// unquoted 00112233 into the number 38043 and 1e5 into 100000. Tagging
+	// a string setting as a string makes the encoder quote such a value.
+	// Other settings stay unquoted, so compression_level 9 is a number.
+	// The encoder refuses to write a value that is not valid UTF-8 as a
+	// string. Left untagged, such a value is written as base64 !!binary and
+	// loads back unchanged.
+	if configKeyIsString(keys) && utf8.ValidString(value) {
+		node.Tag = yamlStringTag
+	}
+
 	return nil
 }
 
-// yamlSetInMapping resolves (creating if needed) the value node for key
-// within a mapping node, setting it to value when it is the final path
-// element, and returns the node to descend into.
-func yamlSetInMapping(node *yaml.Node, key, value string, last bool) *yaml.Node {
+// configKeyIsString reports whether the dotted key names a string in
+// config.Config, following the fields' yaml tags, as s3.access_key_id and
+// snapshots.home.exclude.0 do.
+func configKeyIsString(keys []string) bool {
+	typ := reflect.TypeFor[config.Config]()
+
+	for _, key := range keys {
+		switch {
+		case typ.Kind() == reflect.Map || typ.Kind() == reflect.Slice:
+			// The key is a snapshot name or a list index.
+			typ = typ.Elem()
+		case typ.Kind() == reflect.Struct:
+			field, ok := yamlField(typ, key)
+			if !ok {
+				return false
+			}
+
+			typ = field.Type
+		default:
+			return false
+		}
+	}
+
+	return typ.Kind() == reflect.String
+}
+
+// yamlField returns the field of struct type typ whose yaml tag names key.
+func yamlField(typ reflect.Type, key string) (reflect.StructField, bool) {
+	for field := range typ.Fields() {
+		name, _, _ := strings.Cut(field.Tag.Get("yaml"), ",")
+		if name == key {
+			return field, true
+		}
+	}
+
+	return reflect.StructField{}, false
+}
+
+// yamlSetInMapping resolves (creating if needed) the value node for the
+// path element keys[i] within a mapping node, setting it to value when it
+// is the final path element, and returns the node to descend into.
+func yamlSetInMapping(
+	node *yaml.Node, keys []string, i int, value string, last bool,
+) *yaml.Node {
 	var valueNode *yaml.Node
 
 	for j := 0; j+1 < len(node.Content); j += 2 {
-		if node.Content[j].Value == key {
+		if node.Content[j].Value == keys[i] {
 			valueNode = node.Content[j+1]
 
 			break
@@ -542,13 +772,8 @@ func yamlSetInMapping(node *yaml.Node, key, value string, last bool) *yaml.Node
 	}
 
 	if valueNode == nil {
-		keyNode := &yaml.Node{Kind: yaml.ScalarNode, Value: key}
-
-		valueNode = &yaml.Node{Kind: yaml.MappingNode}
-		if last {
-			valueNode = &yaml.Node{Kind: yaml.ScalarNode, Value: value}
-		}
-
+		keyNode := &yaml.Node{Kind: yaml.ScalarNode, Value: keys[i]}
+		valueNode = newPathNode(keys, i, value)
 		node.Content = append(node.Content, keyNode, valueNode)
 	} else if last {
 		setScalar(valueNode, value)
@@ -576,12 +801,7 @@ func yamlSetInSequence(
 	}
 
 	if idx == len(node.Content) {
-		newNode := &yaml.Node{Kind: yaml.MappingNode}
-		if last {
-			newNode = &yaml.Node{Kind: yaml.ScalarNode, Value: value}
-		}
-
-		node.Content = append(node.Content, newNode)
+		node.Content = append(node.Content, newPathNode(keys, i, value))
 	} else if last {
 		setScalar(node.Content[idx], value)
 	}
@@ -589,6 +809,22 @@ func yamlSetInSequence(
 	return node.Content[idx], nil
 }
 
+// newPathNode returns the node to create for the missing path element
+// keys[i]: the value when it is the final path element, otherwise an empty
+// list when the next path element is a number and an empty map when not.
+func newPathNode(keys []string, i int, value string) *yaml.Node {
+	if i == len(keys)-1 {
+		return &yaml.Node{Kind: yaml.ScalarNode, Value: value}
+	}
+
+	_, err := strconv.Atoi(keys[i+1])
+	if err == nil {
+		return &yaml.Node{Kind: yaml.SequenceNode}
+	}
+
+	return &yaml.Node{Kind: yaml.MappingNode}
+}
+
 // setScalar overwrites a node in place with a plain scalar value.
 func setScalar(n *yaml.Node, value string) {
 	n.Kind = yaml.ScalarNode

internal/cli/config_test.go

@@ -1,11 +1,16 @@
 package cli //nolint:testpackage // exercises unexported yamlPathGet/yamlPathSet
 
 import (
+	"bytes"
+	"os"
+	"path/filepath"
+	"strconv"
 	"strings"
 	"testing"
 
 	"gopkg.in/yaml.v3"
 	"sneak.berlin/go/vaultik/internal/config"
+	"sneak.berlin/go/vaultik/internal/ui"
 )
 
 // TestDefaultConfigTemplateParses ensures the init template is valid YAML
@@ -20,8 +25,10 @@ func TestDefaultConfigTemplateParses(t *testing.T) {
 		t.Fatalf("default config template is not valid YAML: %v", err)
 	}
 
-	if len(cfg.AgeRecipients) != 1 {
-		t.Errorf("expected 1 placeholder age recipient, got %d", len(cfg.AgeRecipients))
+	// A placeholder recipient would fail config.Load, so the template
+	// leaves the list empty.
+	if len(cfg.AgeRecipients) != 0 {
+		t.Errorf("expected no age recipients, got %d", len(cfg.AgeRecipients))
 	}
 
 	home, ok := cfg.Snapshots["home"]
@@ -51,6 +58,353 @@ func TestDefaultConfigTemplateParses(t *testing.T) {
 	}
 }
 
+// testRecipient is a valid age public key.
+const testRecipient = "age1278m9q7dp3chsh2dcy82qk27v047zywyvtxwnj4cvt0z65jw6a7q5dqhfj"
+
+// TestConfigSetRecipientOnFreshConfig follows the README's getting started
+// steps: on the file `config init` writes, `config set age_recipients.0` and
+// `config set storage_url` give a config that loads with that recipient.
+// Neither is refused, although the config does not load until both are set;
+// once it loads, a recipient that is not a valid key is.
+func TestConfigSetRecipientOnFreshConfig(t *testing.T) {
+	t.Parallel()
+
+	path := filepath.Join(t.TempDir(), "config.yml")
+
+	err := os.WriteFile(path, []byte(defaultConfigTemplate), configFileMode)
+	if err != nil {
+		t.Fatalf("write config: %v", err)
+	}
+
+	out := ui.NewWithColor(&bytes.Buffer{}, false)
+
+	err = writeConfigSet(out, path, "age_recipients.0", testRecipient)
+	if err != nil {
+		t.Fatalf("config set age_recipients.0: %v", err)
+	}
+
+	err = writeConfigSet(out, path, "storage_url", "file:///mnt/backups")
+	if err != nil {
+		t.Fatalf("config set storage_url: %v", err)
+	}
+
+	cfg, err := config.Load(path)
+	if err != nil {
+		t.Fatalf("config.Load: %v", err)
+	}
+
+	if len(cfg.AgeRecipients) != 1 || cfg.AgeRecipients[0] != testRecipient {
+		t.Errorf("age_recipients = %v, want [%s]", cfg.AgeRecipients, testRecipient)
+	}
+
+	err = writeConfigSet(out, path, "age_recipients.0", "notakey")
+	if err == nil || !strings.Contains(err.Error(), "not a valid recipient") {
+		t.Errorf("config set age_recipients.0 notakey gives error %v, "+
+			"want one saying it is not a valid recipient", err)
+	}
+
+	_, err = config.Load(path)
+	if err != nil {
+		t.Errorf("config.Load after the refused change: %v", err)
+	}
+}
+
+// configWithoutRecipients loads, but has no age_recipients key and no
+// snapshot named work.
+const configWithoutRecipients = `storage_url: file:///mnt/backups
+snapshots:
+  home:
+    paths:
+      - "~"
+`
+
+// configWithUnsetReference loads only where VAULTIK_TEST_UNSET_SECRET is
+// set, as a backup job's environment might set it. No test sets it.
+const configWithUnsetReference = configWithoutRecipients + `s3:
+  secret_access_key: ${ENV:VAULTIK_TEST_UNSET_SECRET}
+`
+
+// configWithLevelReference returns configWithoutRecipients with a ${FILE:...}
+// reference on compression_level to a file holding 9, so that the config
+// loads once the reference is filled in.
+func configWithLevelReference(t *testing.T) string {
+	t.Helper()
+
+	levelPath := filepath.Join(t.TempDir(), "level")
+
+	err := os.WriteFile(levelPath, []byte("9"), configFileMode)
+	if err != nil {
+		t.Fatalf("write level: %v", err)
+	}
+
+	return configWithoutRecipients +
+		"compression_level: ${FILE:" + levelPath + "}\n"
+}
+
+// TestConfigSetIndexUnderMissingKey checks that a list index under a key
+// the file does not have creates a list, and that the result loads.
+func TestConfigSetIndexUnderMissingKey(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		key   string
+		value string
+		field func(cfg *config.Config) []string
+	}{
+		{"age_recipients.0", testRecipient,
+			func(cfg *config.Config) []string { return cfg.AgeRecipients }},
+		{"snapshots.work.paths.0", "/srv/work",
+			func(cfg *config.Config) []string { return cfg.Snapshots["work"].Paths }},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.key, func(t *testing.T) {
+			t.Parallel()
+
+			path := filepath.Join(t.TempDir(), "config.yml")
+
+			err := os.WriteFile(path, []byte(configWithoutRecipients), configFileMode)
+			if err != nil {
+				t.Fatalf("write config: %v", err)
+			}
+
+			err = writeConfigSet(ui.NewWithColor(&bytes.Buffer{}, false),
+				path, tt.key, tt.value)
+			if err != nil {
+				t.Fatalf("config set %s: %v", tt.key, err)
+			}
+
+			cfg, err := config.Load(path)
+			if err != nil {
+				t.Fatalf("config.Load: %v", err)
+			}
+
+			got := tt.field(cfg)
+			if len(got) != 1 || got[0] != tt.value {
+				t.Errorf("%s gives %v, want [%s]", tt.key, got, tt.value)
+			}
+		})
+	}
+}
+
+// TestConfigSetKeepsFileThatWouldNotLoad checks that config set refuses a
+// change after which the config would not load, with an error giving the
+// reason, and leaves the file as it was. On the file `config init` writes,
+// which does not load until storage_url is set, and on one that cannot be
+// read here, only a value of the wrong type is refused.
+func TestConfigSetKeepsFileThatWouldNotLoad(t *testing.T) {
+	t.Parallel()
+
+	levelReference := configWithLevelReference(t)
+
+	tests := []struct {
+		config string
+		key    string
+		value  string
+		reason string
+	}{
+		// Index 1 of the empty list the missing key gives.
+		{configWithoutRecipients, "age_recipients.1", testRecipient,
+			"index out of range"},
+		// A list set to a single value.
+		{configWithoutRecipients, "snapshots.home.paths", "/srv",
+			"into []string"},
+		{configWithoutRecipients, "chunk_size", "fast",
+			"invalid size format"},
+		// Values of the right type that fail validation.
+		{configWithoutRecipients, "compression_level", "99",
+			"compression_level must be between 1 and 19"},
+		{configWithoutRecipients, "chunk_size", "1KB",
+			"chunk_size must be at least 1MB"},
+		// With its reference filled in, the config loads.
+		{levelReference, "blob_size_limit", "1MB",
+			"blob_size_limit must be at least"},
+		// The type check on a config that does not load yet.
+		{defaultConfigTemplate, "snapshots.apps.paths", "/srv",
+			"into []string"},
+		// The type check on a config that cannot be read here.
+		{configWithUnsetReference, "snapshots.home.paths", "/x",
+			"cannot unmarshal !!str `/x`"},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.key+" "+tt.value, func(t *testing.T) {
+			t.Parallel()
+
+			path := filepath.Join(t.TempDir(), "config.yml")
+
+			err := os.WriteFile(path, []byte(tt.config), configFileMode)
+			if err != nil {
+				t.Fatalf("write config: %v", err)
+			}
+
+			err = writeConfigSet(ui.NewWithColor(&bytes.Buffer{}, false),
+				path, tt.key, tt.value)
+			if err == nil || !strings.Contains(err.Error(), tt.reason) {
+				t.Errorf("config set %s %s gives error %v, want one saying %q",
+					tt.key, tt.value, err, tt.reason)
+			}
+
+			data, err := os.ReadFile(path) //nolint:gosec // G304: test-controlled path
+			if err != nil {
+				t.Fatalf("read config: %v", err)
+			}
+
+			if string(data) != tt.config {
+				t.Errorf("config set %s changed the file to:\n%s",
+					tt.key, data)
+			}
+		})
+	}
+}
+
+// TestConfigSetRefusesOnlyItsOwnFailure checks that setting another key
+// works on a config with a ${...} reference on compression_level, which
+// loads because loading fills the reference in, on one whose
+// compression_level already fails to load, and on one with a reference
+// that cannot be filled in here.
+func TestConfigSetRefusesOnlyItsOwnFailure(t *testing.T) {
+	t.Parallel()
+
+	configs := map[string]string{
+		"reference":       configWithLevelReference(t),
+		"failing":         configWithoutRecipients + "compression_level: fast\n",
+		"unset reference": configWithUnsetReference,
+	}
+
+	for name, configYAML := range configs {
+		t.Run(name, func(t *testing.T) {
+			t.Parallel()
+
+			path := filepath.Join(t.TempDir(), "config.yml")
+
+			err := os.WriteFile(path, []byte(configYAML), configFileMode)
+			if err != nil {
+				t.Fatalf("write config: %v", err)
+			}
+
+			err = writeConfigSet(ui.NewWithColor(&bytes.Buffer{}, false),
+				path, "hostname", "myhost")
+			if err != nil {
+				t.Fatalf("config set hostname: %v", err)
+			}
+
+			data, err := os.ReadFile(path) //nolint:gosec // G304: test-controlled path
+			if err != nil {
+				t.Fatalf("read config: %v", err)
+			}
+
+			if !strings.Contains(string(data), "hostname: myhost") {
+				t.Errorf("config set hostname wrote:\n%s", data)
+			}
+		})
+	}
+}
+
+// TestConfigSetStringLooksLikeNumber sets string settings to values that
+// YAML reads as numbers or booleans when they are unquoted, and checks that
+// config.Load returns each one unchanged.
+func TestConfigSetStringLooksLikeNumber(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		key   string
+		value string
+		field func(cfg *config.Config) string
+	}{
+		{"s3.access_key_id", "00112233",
+			func(cfg *config.Config) string { return cfg.S3.AccessKeyID }},
+		{"s3.secret_access_key", "12345678901234567890123456789012",
+			func(cfg *config.Config) string { return cfg.S3.SecretAccessKey }},
+		{"hostname", "007",
+			func(cfg *config.Config) string { return cfg.Hostname }},
+		{"s3.prefix", "1e5",
+			func(cfg *config.Config) string { return cfg.S3.Prefix }},
+		{"s3.bucket", "true",
+			func(cfg *config.Config) string { return cfg.S3.Bucket }},
+		{"s3.region", "FALSE",
+			func(cfg *config.Config) string { return cfg.S3.Region }},
+		{"snapshots.home.exclude.0", "1.10",
+			func(cfg *config.Config) string { return cfg.Snapshots["home"].Exclude[0] }},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.key+"="+tt.value, func(t *testing.T) {
+			t.Parallel()
+
+			cfg := loadAfterConfigSet(t, tt.key, tt.value)
+
+			got := tt.field(cfg)
+			if got != tt.value {
+				t.Errorf("%s = %q after config set %q", tt.key, got, tt.value)
+			}
+		})
+	}
+}
+
+// TestConfigSetNonUTF8Path checks that config set still accepts a value that
+// is not valid UTF-8, such as a path with a Latin-1 file name, and that
+// config.Load returns it unchanged.
+func TestConfigSetNonUTF8Path(t *testing.T) {
+	t.Parallel()
+
+	const dir = "/srv/caf\xe9"
+
+	cfg := loadAfterConfigSet(t, "snapshots.home.paths.0", dir)
+
+	got := cfg.Snapshots["home"].Paths[0]
+	if got != dir {
+		t.Errorf("snapshots.home.paths.0 = %q, want %q", got, dir)
+	}
+}
+
+// TestConfigSetNumberStaysNumber checks that a number set for an integer
+// setting is still read as a number, not as a quoted string.
+func TestConfigSetNumberStaysNumber(t *testing.T) {
+	t.Parallel()
+
+	const level = 9
+
+	cfg := loadAfterConfigSet(t, "compression_level", strconv.Itoa(level))
+
+	if cfg.CompressionLevel != level {
+		t.Errorf("compression_level = %d, want %d", cfg.CompressionLevel, level)
+	}
+}
+
+// loadAfterConfigSet writes the file `config init` writes, sets storage_url
+// to a local directory so that the file passes validation, applies
+// `config set key value` and returns what config.Load reads back.
+func loadAfterConfigSet(t *testing.T, key, value string) *config.Config {
+	t.Helper()
+
+	path := filepath.Join(t.TempDir(), "config.yml")
+
+	err := os.WriteFile(path, []byte(defaultConfigTemplate), configFileMode)
+	if err != nil {
+		t.Fatalf("write config: %v", err)
+	}
+
+	out := ui.NewWithColor(&bytes.Buffer{}, false)
+
+	err = writeConfigSet(out, path, "storage_url", "file:///mnt/backups")
+	if err != nil {
+		t.Fatalf("config set storage_url: %v", err)
+	}
+
+	err = writeConfigSet(out, path, key, value)
+	if err != nil {
+		t.Fatalf("config set %s: %v", key, err)
+	}
+
+	cfg, err := config.Load(path)
+	if err != nil {
+		t.Fatalf("config.Load: %v", err)
+	}
+
+	return cfg
+}
+
 const testYAML = `# top comment
 compression_level: 3
 age_recipients:
@@ -188,6 +542,135 @@ func TestYAMLPathSet(t *testing.T) {
 	}
 }
 
+// TestConfigSetPreservesFormatting asserts the `config set` write path
+// (marshalConfigYAML) round-trips a 2-space-indented file without reindenting
+// it to yaml.Marshal's 4-space default, and keeps comments.
+func TestConfigSetPreservesFormatting(t *testing.T) {
+	t.Parallel()
+
+	root := parseTestYAML(t)
+
+	err := yamlPathSet(root, splitPath("s3.bucket"), "newbucket")
+	if err != nil {
+		t.Fatalf("set s3.bucket: %v", err)
+	}
+
+	out, err := marshalConfigYAML(root)
+	if err != nil {
+		t.Fatalf("marshal: %v", err)
+	}
+
+	text := string(out)
+
+	for _, want := range []string{"# top comment", "# inline comment"} {
+		if !contains(text, want) {
+			t.Errorf("round-tripped YAML dropped comment %q:\n%s", want, text)
+		}
+	}
+
+	// Nested map keys stay at 2-space indent; the bug reindented them to 4.
+	if !contains(text, "\n  bucket: newbucket") {
+		t.Errorf("expected 2-space indent for s3.bucket, got:\n%s", text)
+	}
+
+	if contains(text, "\n    bucket:") {
+		t.Errorf("s3.bucket reindented to 4 spaces:\n%s", text)
+	}
+
+	// Sequence items under a key also stay at 2 spaces.
+	if !contains(text, "\n  - age1aaa") {
+		t.Errorf("expected 2-space indent for sequence item, got:\n%s", text)
+	}
+}
+
+// TestWriteConfigSetHidesSecret checks that setting a secret key prints
+// only the key name, never the value, to the confirmation output.
+func TestWriteConfigSetHidesSecret(t *testing.T) {
+	t.Parallel()
+
+	const secret = "SUPERSECRETVALUE"
+
+	path := filepath.Join(t.TempDir(), "config.yaml")
+
+	err := os.WriteFile(path, []byte("version: 1\n"), 0o600)
+	if err != nil {
+		t.Fatalf("seed config: %v", err)
+	}
+
+	var buf bytes.Buffer
+
+	err = writeConfigSet(ui.NewWithColor(&buf, false), path,
+		"s3.secret_access_key", secret)
+	if err != nil {
+		t.Fatalf("writeConfigSet: %v", err)
+	}
+
+	if strings.Contains(buf.String(), secret) {
+		t.Errorf("output echoed the secret value: %q", buf.String())
+	}
+
+	if !strings.Contains(buf.String(), "s3.secret_access_key") {
+		t.Errorf("output did not confirm the key name: %q", buf.String())
+	}
+}
+
+// TestWriteConfigSetTightensMode checks that a pre-existing group- or
+// world-readable config is tightened to owner-only after a set, since
+// os.WriteFile leaves an existing file's mode untouched.
+func TestWriteConfigSetTightensMode(t *testing.T) {
+	t.Parallel()
+
+	path := filepath.Join(t.TempDir(), "config.yaml")
+
+	// Seed a world-readable config; the loose mode is the condition under
+	// test, so gosec's G306 is expected here.
+	err := os.WriteFile(path, []byte("version: 1\n"), 0o644) //nolint:gosec // G306
+	if err != nil {
+		t.Fatalf("seed config: %v", err)
+	}
+
+	var buf bytes.Buffer
+
+	err = writeConfigSet(ui.NewWithColor(&buf, false), path,
+		"compression_level", "9")
+	if err != nil {
+		t.Fatalf("writeConfigSet: %v", err)
+	}
+
+	info, err := os.Stat(path)
+	if err != nil {
+		t.Fatalf("stat config: %v", err)
+	}
+
+	if info.Mode().Perm() != 0o600 {
+		t.Errorf("config mode = %04o, want 0600", info.Mode().Perm())
+	}
+}
+
+// TestConfigEditEditorWithArgument checks that an $EDITOR carrying an
+// argument works and that a config path with a space reaches it whole.
+// The "editor" is cp with the replacement file as its argument, so the
+// config ends up with the replacement's content only if both arrived.
+func TestConfigEditEditorWithArgument(t *testing.T) {
+	dir := t.TempDir()
+	path := seedFile(t, dir, "my config.yml", "compression_level: 3\n")
+	edited := seedFile(t, dir, "edited.yml", "compression_level: 9\n")
+
+	setRootFlags(t, RootFlags{ConfigPath: path})
+	t.Setenv("EDITOR", "cp '"+edited+"'")
+
+	mustExecute(t, newConfigEditCommand())
+
+	data, err := os.ReadFile(path) //nolint:gosec // G304: test-controlled path
+	if err != nil {
+		t.Fatalf("reading config back: %v", err)
+	}
+
+	if string(data) != "compression_level: 9\n" {
+		t.Errorf("config after edit = %q, want the edited content", data)
+	}
+}
+
 func splitPath(s string) []string {
 	return strings.Split(s, ".")
 }

internal/cli/database.go

@@ -15,6 +15,12 @@ func NewDatabaseCommand() *cobra.Command {
 		Use:   "database",
 		Short: "Manage the local state database",
 		Long:  `Commands for managing the local SQLite state database.`,
+		// cobra checks Args only on a command that has a Run: without
+		// one, a mistyped subcommand printed help and exited 0.
+		Args: cobra.NoArgs,
+		Run: func(cmd *cobra.Command, _ []string) {
+			_ = cmd.Help()
+		},
 	}
 
 	cmd.AddCommand(
@@ -48,7 +54,7 @@ storage destination on that run.
 
 Use --force to skip the confirmation prompt.`,
 		Args: cobra.NoArgs,
-		RunE: func(_ *cobra.Command, _ []string) error {
+		RunE: func(cmd *cobra.Command, _ []string) error {
 			// Resolve config path
 			configPath, err := ResolveConfigPath()
 			if err != nil {
@@ -62,26 +68,31 @@ Use --force to skip the confirmation prompt.`,
 			}
 
 			dbPath := cfg.IndexPath
+			out := commandUI(cmd)
 
 			// Check if database exists
 			_, err = os.Stat(dbPath)
 			if os.IsNotExist(err) {
-				_, _ = fmt.Fprintf(os.Stdout, "Database does not exist: %s\n", dbPath)
+				out.Infof("Local state database does not exist: %s.", dbPath)
 
 				return nil
 			}
 
-			// Confirm unless --force
+			// Confirm unless --force. The prompt and its immediate result
+			// are an interactive exchange the operator must see, so they go
+			// straight to stdout rather than through the UI and --quiet does
+			// not silence them.
 			if !force {
-				_, _ = fmt.Fprintf(os.Stdout,
+				w := cmd.OutOrStdout()
+				_, _ = fmt.Fprintf(w,
 					"This will delete the local state database at:\n  %s\n\n", dbPath)
-				_, _ = fmt.Fprint(os.Stdout, "Are you sure? Type 'yes' to confirm: ")
+				_, _ = fmt.Fprint(w, "Are you sure? Type 'yes' to confirm: ")
 
 				var confirm string
 
 				_, err = fmt.Scanln(&confirm)
 				if err != nil || confirm != "yes" {
-					_, _ = fmt.Fprintln(os.Stdout, "Aborted.")
+					_, _ = fmt.Fprintln(w, "Aborted.")
 
 					//nolint:nilerr // a failed/aborted confirmation is a clean abort
 					return nil
@@ -100,11 +111,7 @@ Use --force to skip the confirmation prompt.`,
 			_ = os.Remove(walPath) // Ignore errors - files may not exist
 			_ = os.Remove(shmPath)
 
-			rootFlags := GetRootFlags()
-			if !rootFlags.Quiet {
-				_, _ = fmt.Fprintf(os.Stdout, "Database deleted: %s\n", dbPath)
-			}
-
+			out.Infof("Local state database deleted: %s.", dbPath)
 			log.Info("Local state database deleted", "path", dbPath)
 
 			return nil

internal/cli/duration.go

@@ -1,126 +0,0 @@
-package cli
-
-import (
-	"errors"
-	"fmt"
-	"regexp"
-	"strconv"
-	"strings"
-	"time"
-)
-
-// Approximate lengths of the extended calendar units accepted by
-// parseDuration.
-const (
-	durationDay   = 24 * time.Hour
-	durationWeek  = 7 * durationDay
-	durationMonth = 30 * durationDay
-	durationYear  = 365 * durationDay
-)
-
-var (
-	errNegativeDuration = errors.New("negative durations are not supported")
-	errInvalidDuration  = errors.New("invalid duration format")
-	errUnknownTimeUnit  = errors.New("unknown time unit")
-)
-
-// parseDuration parses duration strings. Supports standard Go duration format
-// (e.g., "3h30m", "1h45m30s") as well as extended units:
-// - d: days (e.g., "30d", "7d")
-// - w: weeks (e.g., "2w", "4w")
-// - mo: months (30 days) (e.g., "6mo", "1mo")
-// - y: years (365 days) (e.g., "1y", "2y")
-//
-// Can combine units: "1y6mo", "2w3d", "1d12h30m"
-func parseDuration(s string) (time.Duration, error) {
-	// First try standard Go duration parsing
-	d, err := time.ParseDuration(s)
-	if err == nil {
-		return d, nil
-	}
-
-	// Extended duration parsing
-	// Check for negative values
-	if strings.HasPrefix(strings.TrimSpace(s), "-") {
-		return 0, errNegativeDuration
-	}
-
-	// Pattern matches: number + unit, repeated
-	re := regexp.MustCompile(`(\d+(?:\.\d+)?)\s*([a-zA-Z]+)`)
-	matches := re.FindAllStringSubmatch(s, -1)
-
-	if len(matches) == 0 {
-		return 0, fmt.Errorf("%w: %q", errInvalidDuration, s)
-	}
-
-	var total time.Duration
-
-	for _, match := range matches {
-		valueStr := match[1]
-		unit := strings.ToLower(match[2])
-
-		value, err := strconv.ParseFloat(valueStr, 64)
-		if err != nil {
-			return 0, fmt.Errorf("invalid number %q: %w", valueStr, err)
-		}
-
-		d, err := durationForUnit(value, unit)
-		if err != nil {
-			return 0, err
-		}
-
-		total += d
-	}
-
-	return total, nil
-}
-
-// durationForUnit converts a value with a (case-normalized) unit suffix
-// into a time.Duration, accepting Go's standard units plus the extended
-// calendar units.
-func durationForUnit(value float64, unit string) (time.Duration, error) {
-	switch unit {
-	// Standard time units
-	case "ns", "nanosecond", "nanoseconds":
-		return time.Duration(value), nil
-	case "us", "µs", "microsecond", "microseconds":
-		return time.Duration(value * float64(time.Microsecond)), nil
-	case "ms", "millisecond", "milliseconds":
-		return time.Duration(value * float64(time.Millisecond)), nil
-	case "s", "sec", "second", "seconds":
-		return time.Duration(value * float64(time.Second)), nil
-	case "m", "min", "minute", "minutes":
-		return time.Duration(value * float64(time.Minute)), nil
-	case "h", "hr", "hour", "hours":
-		return time.Duration(value * float64(time.Hour)), nil
-	// Extended units
-	case "d", "day", "days":
-		return time.Duration(value * float64(durationDay)), nil
-	case "w", "week", "weeks":
-		return time.Duration(value * float64(durationWeek)), nil
-	case "mo", "month", "months":
-		// Using 30 days as approximation
-		return time.Duration(value * float64(durationMonth)), nil
-	case "y", "year", "years":
-		// Using 365 days as approximation
-		return time.Duration(value * float64(durationYear)), nil
-	default:
-		// Try parsing as standard Go duration unit
-		testStr := "1" + unit
-
-		_, err := time.ParseDuration(testStr)
-		if err != nil {
-			return 0, fmt.Errorf("%w: %q", errUnknownTimeUnit, unit)
-		}
-
-		// It's a valid Go duration unit, parse the full value
-		fullStr := fmt.Sprintf("%g%s", value, unit)
-
-		d, err := time.ParseDuration(fullStr)
-		if err != nil {
-			return 0, fmt.Errorf("invalid duration %q: %w", fullStr, err)
-		}
-
-		return d, nil
-	}
-}

internal/cli/duration_test.go

@@ -1,299 +0,0 @@
-package cli //nolint:testpackage // needs access to unexported parseDuration
-
-import (
-	"testing"
-	"time"
-
-	"github.com/stretchr/testify/assert"
-	"github.com/stretchr/testify/require"
-)
-
-type parseDurationCase struct {
-	name     string
-	input    string
-	expected time.Duration
-	wantErr  bool
-}
-
-// runParseDurationCases executes a table of parseDuration cases as
-// parallel subtests.
-func runParseDurationCases(t *testing.T, tests []parseDurationCase) {
-	t.Helper()
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			t.Parallel()
-
-			got, err := parseDuration(tt.input)
-
-			if tt.wantErr {
-				require.Error(t, err, "expected error for input %q", tt.input)
-
-				return
-			}
-
-			require.NoError(t, err, "unexpected error for input %q", tt.input)
-			assert.Equal(t, tt.expected, got, "duration mismatch for input %q", tt.input)
-		})
-	}
-}
-
-func TestParseDurationStandard(t *testing.T) {
-	t.Parallel()
-
-	runParseDurationCases(t, []parseDurationCase{
-		{
-			name:     "standard seconds",
-			input:    "30s",
-			expected: 30 * time.Second,
-		},
-		{
-			name:     "standard minutes",
-			input:    "45m",
-			expected: 45 * time.Minute,
-		},
-		{
-			name:     "standard hours",
-			input:    "2h",
-			expected: 2 * time.Hour,
-		},
-		{
-			name:     "standard combined",
-			input:    "3h30m",
-			expected: 3*time.Hour + 30*time.Minute,
-		},
-		{
-			name:     "standard complex",
-			input:    "1h45m30s",
-			expected: 1*time.Hour + 45*time.Minute + 30*time.Second,
-		},
-		{
-			name:     "standard with milliseconds",
-			input:    "1s500ms",
-			expected: 1*time.Second + 500*time.Millisecond,
-		},
-	})
-}
-
-func TestParseDurationExtendedUnits(t *testing.T) {
-	t.Parallel()
-
-	runParseDurationCases(t, []parseDurationCase{
-		// Extended units - days
-		{
-			name:     "single day",
-			input:    "1d",
-			expected: 24 * time.Hour,
-		},
-		{
-			name:     "multiple days",
-			input:    "7d",
-			expected: 7 * 24 * time.Hour,
-		},
-		{
-			name:     "fractional days",
-			input:    "1.5d",
-			expected: 36 * time.Hour,
-		},
-		{
-			name:     "days spelled out",
-			input:    "3days",
-			expected: 3 * 24 * time.Hour,
-		},
-		// Extended units - weeks
-		{
-			name:     "single week",
-			input:    "1w",
-			expected: 7 * 24 * time.Hour,
-		},
-		{
-			name:     "multiple weeks",
-			input:    "4w",
-			expected: 4 * 7 * 24 * time.Hour,
-		},
-		{
-			name:     "weeks spelled out",
-			input:    "2weeks",
-			expected: 2 * 7 * 24 * time.Hour,
-		},
-		// Extended units - months
-		{
-			name:     "single month",
-			input:    "1mo",
-			expected: 30 * 24 * time.Hour,
-		},
-		{
-			name:     "multiple months",
-			input:    "6mo",
-			expected: 6 * 30 * 24 * time.Hour,
-		},
-		{
-			name:     "months spelled out",
-			input:    "3months",
-			expected: 3 * 30 * 24 * time.Hour,
-		},
-		// Extended units - years
-		{
-			name:     "single year",
-			input:    "1y",
-			expected: 365 * 24 * time.Hour,
-		},
-		{
-			name:     "multiple years",
-			input:    "2y",
-			expected: 2 * 365 * 24 * time.Hour,
-		},
-		{
-			name:     "years spelled out",
-			input:    "1year",
-			expected: 365 * 24 * time.Hour,
-		},
-	})
-}
-
-func TestParseDurationCombinedAndErrors(t *testing.T) {
-	t.Parallel()
-
-	runParseDurationCases(t, []parseDurationCase{
-		// Combined extended units
-		{
-			name:     "weeks and days",
-			input:    "2w3d",
-			expected: 2*7*24*time.Hour + 3*24*time.Hour,
-		},
-		{
-			name:     "years and months",
-			input:    "1y6mo",
-			expected: 365*24*time.Hour + 6*30*24*time.Hour,
-		},
-		{
-			name:     "days and hours",
-			input:    "1d12h",
-			expected: 24*time.Hour + 12*time.Hour,
-		},
-		{
-			name:  "complex combination",
-			input: "1y2mo3w4d5h6m7s",
-			expected: 365*24*time.Hour + 2*30*24*time.Hour +
-				3*7*24*time.Hour + 4*24*time.Hour +
-				5*time.Hour + 6*time.Minute + 7*time.Second,
-		},
-		{
-			name:     "with spaces",
-			input:    "1d 12h 30m",
-			expected: 24*time.Hour + 12*time.Hour + 30*time.Minute,
-		},
-		// Edge cases
-		{
-			name:     "zero duration",
-			input:    "0s",
-			expected: 0,
-		},
-		{
-			name:     "large duration",
-			input:    "10y",
-			expected: 10 * 365 * 24 * time.Hour,
-		},
-		// Error cases
-		{
-			name:    "empty string",
-			input:   "",
-			wantErr: true,
-		},
-		{
-			name:    "invalid format",
-			input:   "abc",
-			wantErr: true,
-		},
-		{
-			name:    "unknown unit",
-			input:   "5x",
-			wantErr: true,
-		},
-		{
-			name:    "invalid number",
-			input:   "xyzd",
-			wantErr: true,
-		},
-		{
-			name:    "negative not supported",
-			input:   "-5d",
-			wantErr: true,
-		},
-	})
-}
-
-func TestParseDurationSpecialCases(t *testing.T) {
-	t.Parallel()
-
-	// Test that standard Go durations work exactly as expected
-	standardDurations := []string{
-		"300ms",
-		"1.5h",
-		"2h45m",
-		"72h",
-		"1us",
-		"1µs",
-		"1ns",
-	}
-
-	for _, d := range standardDurations {
-		expected, err := time.ParseDuration(d)
-		require.NoError(t, err)
-
-		got, err := parseDuration(d)
-		require.NoError(t, err)
-		assert.Equal(t, expected, got, "standard duration %q should parse identically", d)
-	}
-}
-
-func TestParseDurationRealWorldExamples(t *testing.T) {
-	t.Parallel()
-
-	// Test real-world snapshot purge scenarios
-	tests := []struct {
-		description string
-		input       string
-		olderThan   time.Duration
-	}{
-		{
-			description: "keep snapshots from last 30 days",
-			input:       "30d",
-			olderThan:   30 * 24 * time.Hour,
-		},
-		{
-			description: "keep snapshots from last 6 months",
-			input:       "6mo",
-			olderThan:   6 * 30 * 24 * time.Hour,
-		},
-		{
-			description: "keep snapshots from last year",
-			input:       "1y",
-			olderThan:   365 * 24 * time.Hour,
-		},
-		{
-			description: "keep snapshots from last week and a half",
-			input:       "1w3d",
-			olderThan:   10 * 24 * time.Hour,
-		},
-		{
-			description: "keep snapshots from last 90 days",
-			input:       "90d",
-			olderThan:   90 * 24 * time.Hour,
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.description, func(t *testing.T) {
-			t.Parallel()
-
-			got, err := parseDuration(tt.input)
-			require.NoError(t, err)
-			assert.Equal(t, tt.olderThan, got)
-
-			// Verify the duration makes sense for snapshot purging
-			assert.Greater(t, got, time.Hour,
-				"snapshot purge duration should be at least an hour")
-		})
-	}
-}

internal/cli/entry.go

@@ -1,6 +1,8 @@
 package cli
 
 import (
+	"errors"
+	"io"
 	"os"
 	"strings"
 	"time"
@@ -13,28 +15,67 @@ import (
 // the startup banner.
 const shortCommitLen = 12
 
-// Entry is the main entry point for the CLI application.
-// It prints the startup banner (unless a quiet flag is present in os.Args),
-// executes the root cobra command, and routes any returned error through
-// the ui.Writer so the user sees a properly formatted "🛑 ERROR:" line.
-func Entry() {
-	if !bannerSuppressedInArgs(os.Args[1:]) {
-		short := globals.Commit
-		if len(short) > shortCommitLen {
-			short = short[:shortCommitLen]
-		}
+// exitCodeInterrupted is the exit status of a command that SIGINT or
+// SIGTERM stopped. It is 128 plus SIGINT's number, 2, which is what a
+// shell reports for a command stopped by Ctrl-C.
+const exitCodeInterrupted = 130
 
-		writeStartupBanner(ui.New(os.Stdout), time.Now().UTC(), short)
-	}
+// Entry is the main entry point for the CLI application.
+// It prints the startup banner to stderr (unless a banner-suppressing
+// flag is present in os.Args — see bannerSuppressedInArgs), executes the
+// root cobra command, and routes any returned error through the
+// ui.Writer so the user sees a properly formatted "🛑 ERROR:" line.
+// The banner goes to stderr because stdout carries only the output the
+// user asked for, such as a completion script or a `config get` value.
+//
+// It returns the process exit code (0 on success, 130 when interrupted,
+// 1 on any other error) rather than calling os.Exit, so that main's
+// deferred profile writers run before the process ends. See run in
+// cmd/vaultik/main.go.
+func Entry() int {
+	emitStartupBanner(os.Args[1:], os.Stderr)
 
 	rootCmd := NewRootCommand()
 	rootCmd.SilenceErrors = true
 
 	err := rootCmd.Execute()
 	if err != nil {
+		// An operation that ran inside the fx app has already reported
+		// its own failure (`snapshot verify --json` puts it in the
+		// document instead); errReported says so. Printing it again
+		// here would double the error line.
+		// Every other error — bad arguments, a config that would not
+		// load, an interrupt — reaches Entry unreported, so it is shown
+		// here.
+		if !errors.Is(err, errReported) {
 			ReportErrorf("%s", err.Error())
-		os.Exit(1)
 		}
+
+		if errors.Is(err, errInterrupted) {
+			return exitCodeInterrupted
+		}
+
+		return 1
+	}
+
+	return 0
+}
+
+// emitStartupBanner writes the startup banner to w unless args (the
+// argument vector with the program name already stripped) contains a
+// flag that suppresses it. Split out of Entry so that the decision is
+// reachable from a test without running the whole CLI.
+func emitStartupBanner(args []string, w io.Writer) {
+	if bannerSuppressedInArgs(args) {
+		return
+	}
+
+	short := globals.Commit
+	if len(short) > shortCommitLen {
+		short = short[:shortCommitLen]
+	}
+
+	writeStartupBanner(ui.New(w), time.Now().UTC(), short)
 }
 
 // ReportErrorf emits a user-facing error to stderr in the standard
@@ -46,9 +87,19 @@ func ReportErrorf(format string, args ...any) {
 }
 
 // bannerSuppressedInArgs reports whether any of args is a flag that
-// should suppress the startup banner (--quiet/-q/--cron). Stops at the
-// "--" argument terminator. Recognizes both long forms and short -q,
-// including combined short flags like "-qv".
+// should suppress the startup banner (--quiet/-q/--cron/--json). Stops
+// at the "--" argument terminator. Recognizes both long forms and short
+// -q, including combined short flags like "-qv".
+//
+// This scans the raw argument vector because the banner is printed
+// before cobra parses anything — deliberately, so that it still appears
+// when cobra rejects the arguments and on --help. The consequence is
+// that a flag is matched wherever it occurs in the vector, including
+// positions where the command it belongs to would not accept it.
+// --json is a subcommand flag rather than a persistent one, but so is
+// --cron (it exists only on `snapshot create`), so this adds no new
+// class of imprecision. The only cost of a false positive is a missing
+// decorative banner.
 func bannerSuppressedInArgs(args []string) bool {
 	for _, a := range args {
 		if a == "--" {
@@ -56,11 +107,13 @@ func bannerSuppressedInArgs(args []string) bool {
 		}
 
 		switch a {
-		case "--quiet", "-q", "--cron":
+		case "--quiet", "-q", "--cron", "--json":
 			return true
 		}
 
-		if strings.HasPrefix(a, "--quiet=") || strings.HasPrefix(a, "--cron=") {
+		if strings.HasPrefix(a, "--quiet=") ||
+			strings.HasPrefix(a, "--cron=") ||
+			strings.HasPrefix(a, "--json=") {
 			return true
 		}
 		// Combined short flags like -qv or -vq.

internal/cli/entry_banner_test.go

@@ -0,0 +1,276 @@
+package cli //nolint:testpackage // needs access to unexported emitStartupBanner
+
+import (
+	"bytes"
+	"encoding/json"
+	"fmt"
+	"io"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/adrg/xdg"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// Command words and flags used to build argument vectors below. They are
+// constants rather than repeated literals so that a rename shows up as a
+// compile error in one place.
+const (
+	cmdSnapshot = "snapshot"
+	cmdList     = "list"
+	cmdCreate   = "create"
+	cmdVerify   = "verify"
+	cmdRemove   = "remove"
+	cmdPrune    = "prune"
+	cmdRemote   = "remote"
+	cmdInfo     = "info"
+
+	flagJSON   = "--json"
+	flagQuiet  = "--quiet"
+	flagConfig = "--config"
+
+	// programName is argv[0] as the real process receives it. Entry
+	// strips it before scanning, so it has to be present.
+	programName = "vaultik"
+
+	// someSnapshotID only fills the positional argument; no test needs
+	// the snapshot to exist.
+	someSnapshotID = "host_2026-01-01T00:00:00Z"
+)
+
+// jsonArgumentVectors are the argument vectors of every --json
+// invocation the CLI accepts, with the program name stripped exactly as
+// Entry strips it. Each one must suppress the banner.
+//
+//nolint:gochecknoglobals // read-only test fixture shared by two tests
+var jsonArgumentVectors = map[string][]string{
+	"snapshot list":   {cmdSnapshot, cmdList, flagJSON},
+	"snapshot verify": {cmdSnapshot, cmdVerify, someSnapshotID, flagJSON},
+	"snapshot remove": {cmdSnapshot, cmdRemove, someSnapshotID, flagJSON},
+	"prune":           {cmdPrune, flagJSON},
+	"remote info":     {cmdRemote, cmdInfo, flagJSON},
+
+	// --json before the subcommand, and with an explicit value: the
+	// scan is positional, so both forms have to be recognized.
+	"json first":      {flagJSON, cmdSnapshot, cmdList},
+	"json with value": {cmdSnapshot, cmdList, flagJSON + "=true"},
+
+	// A --json invocation that also carries a flag with a value, so the
+	// scan cannot be fooled by an argument that consumes the next one.
+	"json with config": {
+		flagConfig, "/nonexistent/vaultik.yml", cmdSnapshot, cmdList, flagJSON,
+	},
+}
+
+// TestJSONInvocationSuppressesBanner checks that every --json
+// invocation suppresses the startup banner, as the README says --json
+// does along with --quiet and --cron. The scan is over the raw argument
+// vector, so each position and spelling of --json is listed.
+func TestJSONInvocationSuppressesBanner(t *testing.T) {
+	t.Parallel()
+
+	for name, argv := range jsonArgumentVectors {
+		t.Run(name, func(t *testing.T) {
+			t.Parallel()
+
+			var banner bytes.Buffer
+
+			emitStartupBanner(argv, &banner)
+
+			assert.Empty(t, banner.String(),
+				"--json suppresses the banner")
+		})
+	}
+}
+
+// TestBannerStillPrintedWithoutSuppressingFlag pins the other half of
+// the contract. Without it, deleting the banner outright would satisfy
+// the test above, and the banner is wanted on interactive invocations.
+func TestBannerStillPrintedWithoutSuppressingFlag(t *testing.T) {
+	t.Parallel()
+
+	for name, argv := range map[string][]string{
+		"no flags": {cmdSnapshot, cmdList},
+		"verbose":  {cmdSnapshot, cmdList, "--verbose"},
+		"after the terminator": {
+			cmdSnapshot, "restore", "--", flagJSON,
+		},
+	} {
+		t.Run(name, func(t *testing.T) {
+			t.Parallel()
+
+			var banner bytes.Buffer
+
+			emitStartupBanner(argv, &banner)
+
+			assert.Contains(t, banner.String(), "starting up at",
+				"the banner belongs on invocations that did not opt out")
+		})
+	}
+}
+
+// TestBannerSuppressedInArgs covers the suppression scan directly,
+// including the flags that suppressed the banner before --json joined
+// them, so that adding --json cannot regress them.
+func TestBannerSuppressedInArgs(t *testing.T) {
+	t.Parallel()
+
+	for name, testCase := range map[string]struct {
+		args       []string
+		suppressed bool
+	}{
+		"quiet long":     {[]string{cmdSnapshot, cmdCreate, flagQuiet}, true},
+		"quiet short":    {[]string{cmdSnapshot, cmdCreate, "-q"}, true},
+		"quiet combined": {[]string{cmdSnapshot, cmdCreate, "-qv"}, true},
+		"cron":           {[]string{cmdSnapshot, cmdCreate, "--cron"}, true},
+		"json":           {[]string{cmdSnapshot, cmdList, flagJSON}, true},
+		"nothing":        {[]string{cmdSnapshot, cmdList}, false},
+		"empty":          {nil, false},
+		"json after dashes": {
+			[]string{cmdSnapshot, cmdList, "--", flagJSON}, false,
+		},
+		"quiet after dashes": {
+			[]string{cmdSnapshot, cmdCreate, "--", "-q"}, false,
+		},
+	} {
+		t.Run(name, func(t *testing.T) {
+			t.Parallel()
+
+			assert.Equal(t, testCase.suppressed,
+				bannerSuppressedInArgs(testCase.args))
+		})
+	}
+}
+
+// hermeticConfig is a complete, valid config that needs no network and
+// no credentials: file:// storage is exempt from the S3 credential
+// checks. A test that lists the destination must create its directory
+// first, because listing a directory that does not exist is an error.
+// Chunk, blob and compression settings are filled in by config.Load.
+const hermeticConfig = `age_recipients:
+  - age1278m9q7dp3chsh2dcy82qk27v047zywyvtxwnj4cvt0z65jw6a7q5dqhfj
+snapshots:
+  test:
+    paths:
+      - %s
+storage_url: file://%s
+index_path: %s
+hostname: test-host
+`
+
+// TestEntryJSONStdoutIsExactlyOneDocument runs the real thing: Entry,
+// with a real argument vector, over the process's real stdout file
+// descriptor, all the way through cobra and the fx graph to the
+// document. It is the assertion the issue asks for — `vaultik snapshot
+// list --json | jq .` with no other flags — with the pipe replaced by a
+// decoder.
+//
+// `snapshot list` is the command chosen because it is the only --json
+// command that reaches its document without a populated destination
+// store: it reads the local index, streams `metadata/` (empty here),
+// and treats an empty destination directory as an empty list rather
+// than a failure.
+//
+// Not parallel: it replaces os.Args, os.Stdout and the xdg globals.
+func TestEntryJSONStdoutIsExactlyOneDocument(t *testing.T) {
+	dir := t.TempDir()
+	configPath := filepath.Join(dir, "config.yml")
+	storeDir := filepath.Join(dir, "store")
+
+	contents := fmt.Sprintf(hermeticConfig,
+		filepath.Join(dir, "source"),
+		storeDir,
+		filepath.Join(dir, "index.sqlite"))
+
+	require.NoError(t,
+		os.WriteFile(configPath, []byte(contents), configFileMode))
+
+	require.NoError(t, os.Mkdir(storeDir, 0o750))
+
+	// The PID lock lives under xdg.DataHome, which xdg resolves at
+	// package init; point it at the temp dir so the test neither
+	// touches nor collides with the real one.
+	t.Setenv("XDG_DATA_HOME", filepath.Join(dir, "data"))
+	xdg.Reload()
+	t.Cleanup(xdg.Reload)
+
+	previousArgs := os.Args
+
+	t.Cleanup(func() {
+		os.Args = previousArgs
+		rootFlags = RootFlags{}
+	})
+
+	os.Args = []string{
+		programName, flagConfig, configPath, cmdSnapshot, cmdList, flagJSON,
+	}
+
+	stdout := captureProcessStdout(t, func() { _ = Entry() })
+
+	requireExactlyOneJSONDocument(t, stdout)
+
+	var snapshots []any
+
+	require.NoError(t, json.Unmarshal([]byte(stdout), &snapshots))
+	assert.Empty(t, snapshots,
+		"a destination store with no snapshots lists none")
+}
+
+// captureProcessStdout redirects the process's own stdout to a pipe for
+// the duration of fn and returns what was written to it. The redirection
+// has to be at the file-descriptor level rather than through an injected
+// writer, because the commands Entry runs reach os.Stdout directly.
+//
+// Not parallel-safe: os.Stdout is process-global.
+func captureProcessStdout(t *testing.T, fn func()) string {
+	t.Helper()
+
+	reader, writer, err := os.Pipe()
+	require.NoError(t, err)
+
+	previous := os.Stdout
+	os.Stdout = writer
+
+	captured := make(chan string, 1)
+
+	go func() {
+		var buf bytes.Buffer
+
+		_, _ = io.Copy(&buf, reader)
+		captured <- buf.String()
+	}()
+
+	fn()
+
+	os.Stdout = previous
+
+	require.NoError(t, writer.Close())
+
+	out := <-captured
+
+	require.NoError(t, reader.Close())
+
+	return out
+}
+
+// requireExactlyOneJSONDocument fails unless stdout decodes as a single
+// JSON value with nothing before or after it — the property that makes
+// `| jq` work.
+func requireExactlyOneJSONDocument(t *testing.T, stdout string) {
+	t.Helper()
+
+	decoder := json.NewDecoder(strings.NewReader(stdout))
+
+	var document any
+
+	err := decoder.Decode(&document)
+	require.NoError(t, err,
+		"stdout must parse as JSON, got:\n%s", stdout)
+
+	_, err = decoder.Token()
+	require.ErrorIs(t, err, io.EOF,
+		"stdout must hold exactly one JSON document, got:\n%s", stdout)
+}

internal/cli/entry_interrupt_test.go

@@ -0,0 +1,203 @@
+package cli //nolint:testpackage // shares runEntry and the argument constants
+
+import (
+	"fmt"
+	"net/http"
+	"net/http/httptest"
+	"os"
+	"os/signal"
+	"path/filepath"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/adrg/xdg"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// stalledStoreConfig is hermeticConfig with an s3:// destination store
+// in place of the file:// one. The server behind it accepts any
+// credentials.
+const stalledStoreConfig = `age_recipients:
+  - age1278m9q7dp3chsh2dcy82qk27v047zywyvtxwnj4cvt0z65jw6a7q5dqhfj
+snapshots:
+  test:
+    paths:
+      - %s
+storage_url: s3://bucket?endpoint=%s&ssl=false
+s3:
+  access_key_id: key
+  secret_access_key: secret
+index_path: %s
+hostname: test-host
+`
+
+// interruptRepeat is how often interruptOnFirstRequest sends SIGINT.
+const interruptRepeat = 50 * time.Millisecond
+
+// TestEntryInterruptedRun sends SIGINT to the test process while a
+// command waits on the destination store, and checks that Entry returns
+// 130 and prints one line on stderr saying the run was interrupted. The
+// store is a local HTTP server that holds every request open, so the
+// command is always mid-operation when the signal arrives. The two
+// cases cover --cron and --json, which silence other output.
+//
+// Not parallel: it signals the process and replaces os.Args, os.Stdout,
+// os.Stderr and the xdg globals.
+//
+//nolint:paralleltest // signals the process and replaces process globals
+func TestEntryInterruptedRun(t *testing.T) {
+	for _, testCase := range []struct {
+		name string
+		args []string
+	}{
+		{
+			name: "snapshot create --cron",
+			args: []string{cmdSnapshot, cmdCreate, "--cron"},
+		},
+		{
+			name: "snapshot verify --json",
+			args: []string{cmdSnapshot, cmdVerify, someSnapshotID, flagJSON},
+		},
+	} {
+		t.Run(testCase.name, func(t *testing.T) {
+			endpoint, requestArrived := startStalledStore(t)
+			configPath := writeStalledStoreConfig(t, endpoint)
+			interruptOnFirstRequest(t, requestArrived)
+
+			code, _, stderr := runEntry(t,
+				append([]string{flagConfig, configPath}, testCase.args...)...)
+
+			assert.Equal(t, 130, code)
+			assert.Equal(t, 1,
+				strings.Count(stderr, errInterrupted.Error()), stderr)
+		})
+	}
+}
+
+// interruptOnFirstRequest sends SIGINT to the test process every
+// interruptRepeat, from the first request to the destination store until
+// the test ends. One signal is not enough: the command can reach the
+// store before fx has started catching signals. The test catches SIGINT
+// too, so that a signal fx is not catching does not kill the test
+// binary.
+func interruptOnFirstRequest(t *testing.T, requestArrived <-chan struct{}) {
+	t.Helper()
+
+	self, err := os.FindProcess(os.Getpid())
+	require.NoError(t, err)
+
+	caught := make(chan os.Signal, 1)
+	signal.Notify(caught, os.Interrupt)
+
+	testEnded := make(chan struct{})
+	senderDone := make(chan struct{})
+
+	// Stop catching SIGINT only after the sender has returned. The sender
+	// waits for each SIGINT it sends to arrive on caught; one still on
+	// its way after signal.Stop would kill the test binary.
+	t.Cleanup(func() {
+		close(testEnded)
+		<-senderDone
+		signal.Stop(caught)
+	})
+
+	go func() {
+		defer close(senderDone)
+
+		select {
+		case <-requestArrived:
+		case <-testEnded:
+			return
+		}
+
+		ticker := time.NewTicker(interruptRepeat)
+		defer ticker.Stop()
+
+		for {
+			// Empty caught, so that the receive below waits for this
+			// SIGINT rather than an earlier one.
+			select {
+			case <-caught:
+			default:
+			}
+
+			sendErr := self.Signal(os.Interrupt)
+			if sendErr != nil {
+				t.Errorf("sending SIGINT: %v", sendErr)
+
+				return
+			}
+
+			<-caught
+
+			select {
+			case <-testEnded:
+				return
+			case <-ticker.C:
+			}
+		}
+	}()
+}
+
+// startStalledStore starts an HTTP server that never answers: each
+// request is held until the client gives up on it or the test ends.
+// It returns the server's host:port and a channel that receives a value
+// when the first request arrives.
+func startStalledStore(t *testing.T) (string, <-chan struct{}) {
+	t.Helper()
+
+	requestArrived := make(chan struct{}, 1)
+	release := make(chan struct{})
+
+	server := httptest.NewServer(http.HandlerFunc(
+		func(_ http.ResponseWriter, r *http.Request) {
+			select {
+			case requestArrived <- struct{}{}:
+			default:
+			}
+
+			select {
+			case <-r.Context().Done():
+			case <-release:
+			}
+		}))
+
+	// Cleanups run last-registered first, so release lets any held
+	// request return before Close waits for it.
+	t.Cleanup(server.Close)
+	t.Cleanup(func() { close(release) })
+
+	return server.Listener.Addr().String(), requestArrived
+}
+
+// writeStalledStoreConfig writes a config whose destination store is the
+// server at endpoint and whose snapshot source holds one small file, so
+// that `snapshot create` has a blob to upload. Returns the config path.
+func writeStalledStoreConfig(t *testing.T, endpoint string) string {
+	t.Helper()
+
+	dir := t.TempDir()
+	configPath := filepath.Join(dir, "config.yml")
+	sourceDir := filepath.Join(dir, "source")
+
+	require.NoError(t, os.Mkdir(sourceDir, 0o750))
+	require.NoError(t, os.WriteFile(filepath.Join(sourceDir, "file.txt"),
+		[]byte("contents"), 0o600))
+
+	contents := fmt.Sprintf(stalledStoreConfig,
+		sourceDir, endpoint, filepath.Join(dir, "index.sqlite"))
+
+	require.NoError(t,
+		os.WriteFile(configPath, []byte(contents), configFileMode))
+
+	// The PID lock lives under xdg.DataHome, which xdg resolves at
+	// package init; point it at the temp dir so the test neither
+	// touches nor collides with the real one.
+	t.Setenv("XDG_DATA_HOME", filepath.Join(dir, "data"))
+	xdg.Reload()
+	t.Cleanup(xdg.Reload)
+
+	return configPath
+}

internal/cli/entry_prune_json_stderr_test.go

@@ -0,0 +1,140 @@
+package cli //nolint:testpackage // shares the prune fixtures and capture helpers
+
+import (
+	"bytes"
+	"io"
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// staleRecordLogMessage is the local-cleanup audit line CleanupLocalSnapshots
+// logs for each stale record. It is exactly the signal issue #112 says a
+// machine consumer lost under --json: gated off stdout, and pinned below
+// the log level on stderr because --json used to force Quiet.
+const staleRecordLogMessage = "Removing stale local snapshot record"
+
+// TestEntryPruneJSONStderrHonoursVerbosity is the end-to-end regression
+// guard for issue #112. Under --json the log level must still follow
+// --verbose/--debug rather than being pinned to WARN, so the
+// local-cleanup records reach stderr under --verbose while stdout stays
+// exactly one JSON document; without --verbose they stay below the
+// level, as they do without --json.
+//
+// Both halves are asserted together on the same run, because the fix has
+// to keep the document clean (issue #108) while freeing stderr.
+//
+// Not parallel: it replaces os.Args, os.Stdout, os.Stderr and the xdg
+// globals.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntryPruneJSONStderrHonoursVerbosity(t *testing.T) {
+	for _, testCase := range []struct {
+		name         string
+		verbose      bool
+		wantOnStderr bool
+	}{
+		{
+			name:         "verbose json surfaces the cleanup record on stderr",
+			verbose:      true,
+			wantOnStderr: true,
+		},
+		{
+			name:         "json alone keeps the cleanup record below the level",
+			verbose:      false,
+			wantOnStderr: false,
+		},
+	} {
+		t.Run(testCase.name, func(t *testing.T) {
+			configPath := writeHermeticPruneConfig(t, true)
+
+			previousArgs := os.Args
+
+			t.Cleanup(func() {
+				os.Args = previousArgs
+				rootFlags = RootFlags{}
+			})
+
+			args := []string{
+				programName, flagConfig, configPath, cmdPrune, flagJSON,
+			}
+			if testCase.verbose {
+				args = append(args, "--verbose")
+			}
+
+			os.Args = args
+
+			stdout, stderr := captureProcessStdoutAndStderr(t,
+				func() { _ = Entry() })
+
+			// The document stays clean in both cases: freeing stderr must
+			// not regress issue #108.
+			requireExactlyOneJSONDocument(t, stdout)
+
+			if testCase.wantOnStderr {
+				assert.Contains(t, stderr, staleRecordLogMessage,
+					"--verbose --json must emit the cleanup record on stderr")
+				assert.Contains(t, stderr, stalePruneSnapshotID,
+					"the record must name the snapshot it removed")
+			} else {
+				assert.NotContains(t, stderr, staleRecordLogMessage,
+					"without --verbose the record stays below the log level")
+			}
+		})
+	}
+}
+
+// captureProcessStdoutAndStderr redirects both of the process's own
+// standard streams to pipes for the duration of fn and returns what was
+// written to each. The redirection is at the file-descriptor level
+// because the logger binds os.Stderr when it initializes inside fn, and
+// the JSON document reaches os.Stdout independently; the point is to see
+// where each actually lands.
+//
+// Not parallel-safe: os.Stdout and os.Stderr are process-global.
+func captureProcessStdoutAndStderr(t *testing.T, fn func()) (string, string) {
+	t.Helper()
+
+	outReader, outWriter, err := os.Pipe()
+	require.NoError(t, err)
+
+	errReader, errWriter, err := os.Pipe()
+	require.NoError(t, err)
+
+	previousOut, previousErr := os.Stdout, os.Stderr
+	os.Stdout, os.Stderr = outWriter, errWriter
+
+	capturedOut := drain(outReader)
+	capturedErr := drain(errReader)
+
+	fn()
+
+	os.Stdout, os.Stderr = previousOut, previousErr
+
+	require.NoError(t, outWriter.Close())
+	require.NoError(t, errWriter.Close())
+
+	out, errOut := <-capturedOut, <-capturedErr
+
+	require.NoError(t, outReader.Close())
+	require.NoError(t, errReader.Close())
+
+	return out, errOut
+}
+
+// drain copies a reader to a string on a goroutine and delivers the
+// result once the writer end is closed.
+func drain(reader io.Reader) <-chan string {
+	captured := make(chan string, 1)
+
+	go func() {
+		var buf bytes.Buffer
+
+		_, _ = io.Copy(&buf, reader)
+		captured <- buf.String()
+	}()
+
+	return captured
+}

internal/cli/entry_prune_json_test.go

@@ -0,0 +1,176 @@
+package cli //nolint:testpackage // shares hermeticConfig and the capture helpers
+
+import (
+	"context"
+	"database/sql"
+	"encoding/json"
+	"fmt"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/adrg/xdg"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/database"
+	"sneak.berlin/go/vaultik/internal/types"
+)
+
+// pruneJSONDocument is the shape `prune --json` writes: the
+// PruneBlobsResult document, and nothing else.
+//
+//nolint:tagliatelle // snake_case is the established JSON output format
+type pruneJSONDocument struct {
+	BlobsFound   int   `json:"blobs_found"`
+	BlobsDeleted int   `json:"blobs_deleted"`
+	BytesFreed   int64 `json:"bytes_freed"`
+}
+
+// stalePruneSnapshotID is seeded into the local index with no manifest
+// on the destination store, which is exactly what makes it stale.
+const stalePruneSnapshotID = "test-host_test_2026-04-01T09:00:00Z"
+
+// TestEntryPruneJSONStdoutIsExactlyOneDocument is the end-to-end
+// regression guard for issue #108: `vaultik prune --json | jq .` must
+// work with no other flags.
+//
+// It runs Entry over the process's real stdout descriptor, through
+// cobra and the fx graph, against a hermetic file:// destination store
+// — the same construction TestEntryJSONStdoutIsExactlyOneDocument uses
+// for `snapshot list`, with the pipe to jq replaced by a decoder.
+//
+// Both branches of the local-snapshot reconciliation are exercised
+// because the three stdout writes that broke this covered all of them:
+// one line per stale record and a summary when there were any, and a
+// "No stale local snapshots found." line when there were none. No input
+// avoided the contamination, so no single branch demonstrates the fix.
+//
+// Not parallel: it replaces os.Args, os.Stdout and the xdg globals.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout and the xdg globals
+func TestEntryPruneJSONStdoutIsExactlyOneDocument(t *testing.T) {
+	for _, testCase := range []struct {
+		name        string
+		seedStale   bool
+		description string
+	}{
+		{
+			name:        "no stale local records",
+			seedStale:   false,
+			description: "the empty-index branch used to print a 'No stale' line",
+		},
+		{
+			name:      "stale local records present",
+			seedStale: true,
+			description: "the removal branch used to print a line per record " +
+				"plus a summary",
+		},
+	} {
+		t.Run(testCase.name, func(t *testing.T) {
+			configPath := writeHermeticPruneConfig(t, testCase.seedStale)
+
+			previousArgs := os.Args
+
+			t.Cleanup(func() {
+				os.Args = previousArgs
+				rootFlags = RootFlags{}
+			})
+
+			os.Args = []string{
+				programName, flagConfig, configPath, cmdPrune, flagJSON,
+			}
+
+			stdout := captureProcessStdout(t, func() { _ = Entry() })
+
+			requireExactlyOneJSONDocument(t, stdout)
+
+			var document pruneJSONDocument
+
+			require.NoError(t, json.Unmarshal([]byte(stdout), &document),
+				testCase.description)
+
+			// A destination store with no blobs has none to prune. The
+			// assertion that matters is the one above; this one keeps the
+			// test honest about which document it decoded.
+			assert.Equal(t, 0, document.BlobsFound)
+		})
+	}
+}
+
+// writeHermeticPruneConfig builds a config over a temp directory with an
+// empty destination directory and, if seedStale is set, creates the
+// index database up front with one snapshot record that has no
+// counterpart on the destination store, and gives the destination the
+// marker a backup writes. Returns the config path.
+func writeHermeticPruneConfig(t *testing.T, seedStale bool) string {
+	t.Helper()
+
+	dir := t.TempDir()
+	configPath := filepath.Join(dir, "config.yml")
+	indexPath := filepath.Join(dir, "index.sqlite")
+	storeDir := filepath.Join(dir, "store")
+
+	contents := fmt.Sprintf(hermeticConfig,
+		filepath.Join(dir, "source"),
+		storeDir,
+		indexPath)
+
+	require.NoError(t,
+		os.WriteFile(configPath, []byte(contents), configFileMode))
+
+	// prune fails on a destination directory that does not exist, so the
+	// empty store is created here rather than left to a first backup.
+	require.NoError(t, os.Mkdir(storeDir, 0o750))
+
+	// The PID lock lives under xdg.DataHome, which xdg resolves at
+	// package init; point it at the temp dir so the test neither
+	// touches nor collides with the real one.
+	t.Setenv("XDG_DATA_HOME", filepath.Join(dir, "data"))
+	xdg.Reload()
+	t.Cleanup(xdg.Reload)
+
+	if seedStale {
+		seedStaleSnapshotRecord(t, indexPath)
+
+		// While the local index holds a snapshot, prune fails on a
+		// destination that answers that it has no marker.
+		require.NoError(t, os.WriteFile(
+			filepath.Join(storeDir, "vaultik-destination.txt"), nil, 0o600))
+	}
+
+	return configPath
+}
+
+// seedStaleSnapshotRecord creates the index database at path and
+// inserts one completed snapshot into it. Nothing is written to the
+// destination store, so `prune` finds the record stale and removes it —
+// the branch that printed a line per record.
+func seedStaleSnapshotRecord(t *testing.T, path string) {
+	t.Helper()
+
+	ctx := context.Background()
+
+	db, err := database.New(ctx, path)
+	require.NoError(t, err)
+
+	defer func() { require.NoError(t, db.Close()) }()
+
+	startedAt := time.Date(2026, 4, 1, 9, 0, 0, 0, time.UTC)
+	completedAt := startedAt.Add(time.Minute)
+
+	snap := &database.Snapshot{
+		ID:             types.SnapshotID(stalePruneSnapshotID),
+		Hostname:       "test-host",
+		VaultikVersion: "test",
+		StartedAt:      startedAt,
+		CompletedAt:    &completedAt,
+	}
+
+	repos := database.NewRepositories(db)
+
+	err = repos.WithTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+		return repos.Snapshots.Create(ctx, tx, snap)
+	})
+	require.NoError(t, err)
+}

internal/cli/entry_prune_quiet_test.go

@@ -0,0 +1,127 @@
+package cli //nolint:testpackage // shares runEntry and the hermetic config helpers
+
+import (
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// flagCron is used to build the argument vectors below.
+const flagCron = "--cron"
+
+// plantUnreferencedBlob writes a blob that no snapshot references to the
+// destination store at storeDir, dated long before any prune grace period,
+// so that a prune deletes it. Returns its path.
+func plantUnreferencedBlob(t *testing.T, storeDir string) string {
+	t.Helper()
+
+	hash := strings.Repeat("b", 64)
+	path := filepath.Join(storeDir, "blobs", hash[:2], hash[2:4], hash)
+
+	require.NoError(t, os.MkdirAll(filepath.Dir(path), 0o750))
+	require.NoError(t, os.WriteFile(path, []byte("unreferenced"), 0o600))
+
+	longAgo := time.Unix(0, 0)
+	require.NoError(t, os.Chtimes(path, longAgo, longAgo))
+
+	return path
+}
+
+// TestEntrySnapshotCreatePruneIsSilent checks that a successful
+// `snapshot create --prune` under --cron or --quiet prints nothing on
+// stdout or stderr, whether or not the prune finds a blob to delete.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntrySnapshotCreatePruneIsSilent(t *testing.T) {
+	for _, testCase := range []struct {
+		name      string
+		args      []string
+		plantBlob bool
+	}{
+		{
+			name: "cron, no blob to delete",
+			args: []string{cmdSnapshot, cmdCreate, flagCron, flagPrune},
+		},
+		{
+			name:      "cron, blob to delete",
+			args:      []string{cmdSnapshot, cmdCreate, flagCron, flagPrune},
+			plantBlob: true,
+		},
+		{
+			name: "cron with keep-newer-than",
+			args: []string{
+				cmdSnapshot, cmdCreate, flagCron, flagPrune,
+				flagKeepNewerThan, "4w",
+			},
+			plantBlob: true,
+		},
+		{
+			name:      "quiet",
+			args:      []string{flagQuiet, cmdSnapshot, cmdCreate, flagPrune},
+			plantBlob: true,
+		},
+	} {
+		t.Run(testCase.name, func(t *testing.T) {
+			configPath, storeDir := writeBackupConfig(t)
+
+			var blob string
+			if testCase.plantBlob {
+				blob = plantUnreferencedBlob(t, storeDir)
+			}
+
+			code, stdout, stderr := runEntry(t,
+				append([]string{flagConfig, configPath}, testCase.args...)...)
+
+			assert.Equal(t, 0, code)
+			assert.Empty(t, stdout)
+			assert.Empty(t, stderr)
+
+			if testCase.plantBlob {
+				assert.NoFileExists(t, blob, "the prune must delete the blob")
+			}
+		})
+	}
+}
+
+// TestEntryPruneQuietIsSilent checks that a successful `prune --force`
+// under --quiet prints nothing on stdout or stderr, both when there is
+// nothing to remove and when it removes a stale local snapshot record
+// and an unreferenced blob.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntryPruneQuietIsSilent(t *testing.T) {
+	for _, testCase := range []struct {
+		name               string
+		staleRecordAndBlob bool
+	}{
+		{name: "nothing to remove"},
+		{name: "stale record and blob to remove", staleRecordAndBlob: true},
+	} {
+		t.Run(testCase.name, func(t *testing.T) {
+			configPath := writeHermeticPruneConfig(t, testCase.staleRecordAndBlob)
+
+			var blob string
+			if testCase.staleRecordAndBlob {
+				// writeHermeticPruneConfig puts the store beside the config.
+				blob = plantUnreferencedBlob(t,
+					filepath.Join(filepath.Dir(configPath), "store"))
+			}
+
+			code, stdout, stderr := runEntry(t,
+				flagConfig, configPath, flagQuiet, cmdPrune, flagForce)
+
+			assert.Equal(t, 0, code)
+			assert.Empty(t, stdout)
+			assert.Empty(t, stderr)
+
+			if testCase.staleRecordAndBlob {
+				assert.NoFileExists(t, blob, "the prune must delete the blob")
+			}
+		})
+	}
+}

internal/cli/entry_snapshot_create_prune_test.go

@@ -0,0 +1,185 @@
+package cli //nolint:testpackage // shares runEntry and the hermetic config helpers
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/database"
+)
+
+// Flags of `snapshot create` used to build argument vectors below.
+const (
+	flagPrune         = "--prune"
+	flagKeepNewerThan = "--keep-newer-than"
+)
+
+// TestEntrySnapshotCreateRejectsKeepNewerThanBeforeBackup checks that a
+// --keep-newer-than the post-backup prune could not use stops `snapshot
+// create` before the backup: the command fails and the destination
+// store stays empty. It used to be read only by the prune, after the
+// snapshot was uploaded, and without --prune it was ignored.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntrySnapshotCreateRejectsKeepNewerThanBeforeBackup(t *testing.T) {
+	for _, testCase := range []struct {
+		name string
+		args []string
+	}{
+		{
+			name: "decimal in a day count",
+			args: []string{flagPrune, flagKeepNewerThan, "1.5d"},
+		},
+		{
+			name: "unknown unit",
+			args: []string{flagPrune, flagKeepNewerThan, "4x"},
+		},
+		{
+			name: "without --prune",
+			args: []string{flagKeepNewerThan, "4w"},
+		},
+	} {
+		t.Run(testCase.name, func(t *testing.T) {
+			configPath, storeDir := writeBackupConfig(t)
+
+			code, _, _ := runEntry(t, append(
+				[]string{flagConfig, configPath, cmdSnapshot, cmdCreate},
+				testCase.args...)...)
+
+			assert.Equal(t, 1, code)
+
+			stored, err := os.ReadDir(storeDir)
+			require.NoError(t, err)
+			assert.Empty(t, stored, "nothing may be uploaded")
+		})
+	}
+}
+
+// TestEntrySnapshotCreateReportsPostBackupPruneFailure makes the
+// post-backup prune fail after a good backup, by planting a manifest
+// that cannot be decoded, and checks that the failure is reported as the
+// prune's and not as a failed snapshot creation.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntrySnapshotCreateReportsPostBackupPruneFailure(t *testing.T) {
+	configPath, storeDir := writeBackupConfig(t)
+
+	// A remote key is 64 lowercase hex characters. The prune reads every
+	// manifest on the store and fails on one it cannot decode.
+	plantedDir := filepath.Join(storeDir, "metadata", strings.Repeat("a", 64))
+	require.NoError(t, os.MkdirAll(plantedDir, 0o750))
+	require.NoError(t, os.WriteFile(
+		filepath.Join(plantedDir, "manifest.json.zst"),
+		[]byte("not a valid manifest"), 0o600))
+
+	code, _, stderr := runEntry(t,
+		flagConfig, configPath, cmdSnapshot, cmdCreate, flagPrune)
+
+	assert.Equal(t, 1, code)
+	assert.Contains(t, stderr, "Backup completed, but post-backup prune failed")
+	assert.NotContains(t, stderr, "Snapshot creation failed")
+
+	snapshots, err := os.ReadDir(filepath.Join(storeDir, "metadata"))
+	require.NoError(t, err)
+	assert.Len(t, snapshots, 2,
+		"the new snapshot is on the store beside the planted one")
+}
+
+// TestEntrySnapshotCreatePruneExitsNonZeroWhenSourceIsEmpty backs up,
+// empties the source directory, as a volume that is not mounted leaves
+// its mount point, and runs `snapshot create --prune`. The command exits
+// non-zero and reports the backup as completed, not as a failed snapshot
+// creation. See https://git.eeqj.de/sneak/vaultik/issues/388.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntrySnapshotCreatePruneExitsNonZeroWhenSourceIsEmpty(t *testing.T) {
+	configPath, storeDir := writeBackupConfig(t)
+
+	code, _, _ := runEntry(t, flagConfig, configPath, cmdSnapshot, cmdCreate)
+	require.Equal(t, 0, code)
+
+	// writeBackupConfig puts the source directory beside the config file.
+	require.NoError(t, os.Remove(
+		filepath.Join(filepath.Dir(configPath), "source", "file.txt")))
+
+	code, _, stderr := runEntry(t,
+		flagConfig, configPath, cmdSnapshot, cmdCreate, flagPrune)
+
+	assert.Equal(t, 1, code)
+	assert.Contains(t, stderr,
+		"Backup completed, but 1 source path(s) were empty")
+	assert.NotContains(t, stderr, "Snapshot creation failed")
+
+	snapshots, err := os.ReadDir(filepath.Join(storeDir, "metadata"))
+	require.NoError(t, err)
+	assert.Len(t, snapshots, 2, "the older snapshot is kept")
+}
+
+// TestEntrySnapshotCreatePruneExitsNonZeroWhenMountPointIsMissing backs
+// up, records in the local index that this backup found the source
+// directory as a mount point, and runs `snapshot create --prune`, which
+// does not find it as one. The command exits non-zero, reports the
+// backup as completed, not as a failed snapshot creation, and keeps the
+// older snapshot. See https://git.eeqj.de/sneak/vaultik/issues/395.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntrySnapshotCreatePruneExitsNonZeroWhenMountPointIsMissing(
+	t *testing.T,
+) {
+	configPath, storeDir := writeBackupConfig(t)
+
+	code, _, _ := runEntry(t, flagConfig, configPath, cmdSnapshot, cmdCreate)
+	require.Equal(t, 0, code)
+
+	// writeBackupConfig puts the source directory and the local index
+	// beside the config file.
+	dir := filepath.Dir(configPath)
+	ctx := context.Background()
+
+	db, err := database.New(ctx, filepath.Join(dir, "index.sqlite"))
+	require.NoError(t, err)
+
+	repos := database.NewRepositories(db)
+
+	backedUp, err := repos.Snapshots.GetOnlySnapshot(ctx)
+	require.NoError(t, err)
+	require.NoError(t, repos.Snapshots.AddMountPoint(ctx, nil,
+		backedUp.ID.String(), filepath.Join(dir, "source")))
+	require.NoError(t, db.Close())
+
+	code, _, stderr := runEntry(t,
+		flagConfig, configPath, cmdSnapshot, cmdCreate, flagPrune)
+
+	assert.Equal(t, 1, code)
+	assert.Contains(t, stderr, "Backup completed, but 1 mount point(s) "+
+		"an earlier snapshot recorded were not found")
+	assert.NotContains(t, stderr, "Snapshot creation failed")
+
+	snapshots, err := os.ReadDir(filepath.Join(storeDir, "metadata"))
+	require.NoError(t, err)
+	assert.Len(t, snapshots, 2, "the older snapshot is kept")
+}
+
+// writeBackupConfig builds a hermetic config with one small file in its
+// source directory and an empty destination store directory. Returns the
+// config path and the store directory.
+func writeBackupConfig(t *testing.T) (string, string) {
+	t.Helper()
+
+	configPath := writeHermeticPruneConfig(t, false)
+
+	// writeHermeticPruneConfig puts the source and store directories
+	// beside the config file.
+	dir := filepath.Dir(configPath)
+	sourceDir := filepath.Join(dir, "source")
+
+	require.NoError(t, os.Mkdir(sourceDir, 0o750))
+	require.NoError(t, os.WriteFile(
+		filepath.Join(sourceDir, "file.txt"), []byte("hello"), 0o600))
+
+	return configPath, filepath.Join(dir, "store")
+}

internal/cli/entry_snapshot_create_skip_errors_test.go

@@ -0,0 +1,30 @@
+package cli //nolint:testpackage // shares runEntry and the hermetic config helpers
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+// TestEntrySnapshotCreateSkipErrorsExitsNonZeroWhenItSkips backs up,
+// under --skip-errors, a source directory that does not exist, which the
+// walk skips. The command exits non-zero and reports the backup as
+// completed, not as a failed snapshot creation. See
+// https://git.eeqj.de/sneak/vaultik/issues/310.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntrySnapshotCreateSkipErrorsExitsNonZeroWhenItSkips(t *testing.T) {
+	// writeBackupConfig would create the source directory.
+	configPath := writeHermeticPruneConfig(t, false)
+
+	code, stdout, stderr := runEntry(t,
+		flagConfig, configPath, "--skip-errors", cmdSnapshot, cmdCreate)
+
+	assert.Equal(t, 1, code)
+	assert.Contains(t, stdout,
+		"Files: 1 examined, 0 backed up, 0 unchanged, 1 failed.")
+	assert.NotContains(t, stdout, "Finished successfully.")
+	assert.Contains(t, stderr,
+		"Backup completed, but 1 file(s) could not be read")
+	assert.NotContains(t, stderr, "Snapshot creation failed")
+}

internal/cli/entry_snapshot_purge_test.go

@@ -0,0 +1,75 @@
+package cli //nolint:testpackage // shares runEntry and the hermetic config helpers
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/database"
+	"sneak.berlin/go/vaultik/internal/snapshot"
+)
+
+// Words and flags of `snapshot purge` used to build argument vectors
+// below.
+const (
+	cmdPurge      = "purge"
+	flagOlderThan = "--older-than"
+	flagForce     = "--force"
+)
+
+// TestEntrySnapshotPurgeKeepsMetadataWhenLocalIndexDeleteFails makes
+// the local index refuse to delete an old snapshot and checks that
+// `snapshot purge` exits non-zero and leaves that snapshot's metadata on
+// the destination store. It used to log the error, delete the metadata
+// and exit 0.
+//
+// `snapshot create --prune` runs the same purge, but cannot be driven
+// this way: the refusal also breaks the backup's metadata export, which
+// deletes the other snapshots from its copy of the local index.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntrySnapshotPurgeKeepsMetadataWhenLocalIndexDeleteFails(t *testing.T) {
+	configPath := writeHermeticPruneConfig(t, false)
+	dir := filepath.Dir(configPath)
+	indexPath := filepath.Join(dir, "index.sqlite")
+
+	// seedStaleSnapshotRecord writes an old snapshot to the local index
+	// only; the manifest planted here puts it on the destination store
+	// too, so the purge selects it rather than dropping it as stale.
+	seedStaleSnapshotRecord(t, indexPath)
+
+	// While the local index holds a snapshot, purge fails on a
+	// destination that answers that it has no marker.
+	require.NoError(t, os.WriteFile(
+		filepath.Join(dir, "store", "vaultik-destination.txt"), nil, 0o600))
+
+	manifestDir := filepath.Join(dir, "store", "metadata",
+		snapshot.RemoteSnapshotKey(stalePruneSnapshotID))
+	require.NoError(t, os.MkdirAll(manifestDir, 0o750))
+
+	manifestPath := filepath.Join(manifestDir, "manifest.json.zst")
+	require.NoError(t, os.WriteFile(manifestPath, []byte("stub"), 0o600))
+
+	ctx := context.Background()
+
+	db, err := database.New(ctx, indexPath)
+	require.NoError(t, err)
+
+	_, err = db.Conn().ExecContext(ctx, `
+		CREATE TRIGGER refuse_snapshot_delete BEFORE DELETE ON snapshots
+		BEGIN SELECT RAISE(ABORT, 'simulated index error'); END`)
+	require.NoError(t, err)
+	require.NoError(t, db.Close())
+
+	code, _, stderr := runEntry(t, flagConfig, configPath,
+		cmdSnapshot, cmdPurge, flagOlderThan, "30d", flagForce)
+
+	assert.Equal(t, 1, code)
+	assert.Contains(t, stderr, "simulated index error",
+		"the purge must fail at the local-index delete")
+	assert.FileExists(t, manifestPath,
+		"the snapshot is still in the local index, so its metadata must stay")
+}

internal/cli/entry_status_test.go

@@ -0,0 +1,58 @@
+package cli //nolint:testpackage // shares programName and the capture helpers
+
+import (
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+// TestEntryReturnsStatusCode pins the contract main() relies on for
+// issue #75: Entry reports success or failure through its return value
+// and never calls os.Exit. An os.Exit from inside Entry would skip
+// main's deferred profile writers and truncate the profile of a failing
+// command. main turns this code into os.Exit only after those defers
+// run, so a failing command must come back with a non-zero code rather
+// than ending the process here.
+//
+// Stdout and stderr are captured only to keep the banner and command
+// output off the test log; the assertion is on the returned code.
+//
+//nolint:paralleltest // replaces os.Args and rootFlags
+func TestEntryReturnsStatusCode(t *testing.T) {
+	for _, testCase := range []struct {
+		name string
+		args []string
+		want int
+	}{
+		{
+			// version is self-contained: it needs no config and no
+			// destination store, so it exercises the success path.
+			name: "successful command returns zero",
+			args: []string{programName, "version"},
+			want: 0,
+		},
+		{
+			name: "unknown command returns one",
+			args: []string{programName, "no-such-command"},
+			want: 1,
+		},
+	} {
+		t.Run(testCase.name, func(t *testing.T) {
+			previousArgs := os.Args
+
+			t.Cleanup(func() {
+				os.Args = previousArgs
+				rootFlags = RootFlags{}
+			})
+
+			os.Args = testCase.args
+
+			var code int
+
+			_, _ = captureProcessStdoutAndStderr(t, func() { code = Entry() })
+
+			assert.Equal(t, testCase.want, code)
+		})
+	}
+}

internal/cli/entry_stdout_stderr_test.go

@@ -0,0 +1,200 @@
+package cli //nolint:testpackage // shares hermeticConfig and the capture helpers
+
+import (
+	"context"
+	"encoding/json"
+	"fmt"
+	"log/slog"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/adrg/xdg"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"sneak.berlin/go/vaultik/internal/database"
+)
+
+// TestEntryCompletionStdoutIsTheScript runs `vaultik completion bash`,
+// whose stdout the README tells the user to source. The script has to
+// start on the first line.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout and os.Stderr
+func TestEntryCompletionStdoutIsTheScript(t *testing.T) {
+	code, stdout, _ := runEntry(t, "completion", "bash")
+
+	require.Equal(t, 0, code)
+
+	firstLine, _, _ := strings.Cut(stdout, "\n")
+	assert.True(t, strings.HasPrefix(firstLine, "# bash completion"),
+		"the first line of stdout must be the script's, got %q", firstLine)
+}
+
+// TestEntryConfigGetStdoutIsTheValue runs `vaultik config get`, whose
+// stdout a script reads as the value and nothing else.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout and os.Stderr
+func TestEntryConfigGetStdoutIsTheValue(t *testing.T) {
+	configPath := filepath.Join(t.TempDir(), "config.yml")
+	require.NoError(t, os.WriteFile(configPath,
+		[]byte("hostname: test-host\n"), configFileMode))
+
+	code, stdout, _ := runEntry(t,
+		flagConfig, configPath, "config", "get", "hostname")
+
+	require.Equal(t, 0, code)
+	assert.Equal(t, "test-host\n", stdout)
+}
+
+// TestEntryJSONFailureIsReportedOnStderr runs each --json command that
+// writes no document when it fails, against a destination it cannot
+// use. The error must reach stderr, and stdout must stay empty.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntryJSONFailureIsReportedOnStderr(t *testing.T) {
+	for _, testCase := range []struct {
+		name         string
+		args         []string
+		wantOnStderr string
+	}{
+		{
+			name:         "remote info",
+			args:         []string{cmdRemote, cmdInfo, flagJSON},
+			wantOnStderr: "Failed to get remote info",
+		},
+		{
+			name:         "prune",
+			args:         []string{cmdPrune, flagJSON},
+			wantOnStderr: "Prune failed",
+		},
+		{
+			name:         "snapshot remove",
+			args:         []string{cmdSnapshot, cmdRemove, someSnapshotID, flagJSON},
+			wantOnStderr: "Failed to remove snapshot",
+		},
+	} {
+		t.Run(testCase.name, func(t *testing.T) {
+			configPath := writeUnusableDestinationConfig(t)
+
+			code, stdout, stderr := runEntry(t,
+				append([]string{flagConfig, configPath}, testCase.args...)...)
+
+			assert.Equal(t, 1, code)
+			assert.Empty(t, stdout,
+				"a failed --json command has no document to write")
+			assert.Contains(t, stderr, testCase.wantOnStderr,
+				"the failure must be reported on stderr")
+		})
+	}
+}
+
+// TestEntrySnapshotRemoveJSONWarningIsOnStderr runs `snapshot remove
+// --json` on a snapshot in the local index, against a destination
+// directory that does not exist. The command removes the snapshot from
+// the local index and still exits 0. Its stdout must hold the document
+// alone, with the warning about the destination store on stderr: the
+// command to run again once it is reachable, and the snapshot's ID in
+// the record's snapshot_id field.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout, os.Stderr and the xdg globals
+func TestEntrySnapshotRemoveJSONWarningIsOnStderr(t *testing.T) {
+	configPath, indexPath := writeMissingDestinationConfig(t)
+	seedStaleSnapshotRecord(t, indexPath)
+
+	code, stdout, stderr := runEntry(t, flagConfig, configPath,
+		cmdSnapshot, cmdRemove, stalePruneSnapshotID, flagJSON)
+
+	require.Equal(t, 0, code)
+	requireExactlyOneJSONDocument(t, stdout)
+
+	// stderr is a pipe here, so the logger writes one JSON record a line.
+	var warning map[string]any
+
+	for line := range strings.Lines(stderr) {
+		if strings.Contains(line,
+			"Could not remove snapshot metadata from remote storage") {
+			require.NoError(t, json.Unmarshal([]byte(line), &warning))
+		}
+	}
+
+	require.NotNil(t, warning, "the warning must reach stderr")
+	assert.Contains(t, warning[slog.MessageKey],
+		"run 'vaultik snapshot remove' with the snapshot's ID again")
+	assert.Equal(t, stalePruneSnapshotID, warning["snapshot_id"])
+}
+
+// writeMissingDestinationConfig builds a config whose destination
+// directory does not exist. Returns the config path and the path of
+// its local index, which is not created here.
+func writeMissingDestinationConfig(t *testing.T) (string, string) {
+	t.Helper()
+
+	dir := t.TempDir()
+	configPath := filepath.Join(dir, "config.yml")
+	indexPath := filepath.Join(dir, "index.sqlite")
+
+	contents := fmt.Sprintf(hermeticConfig,
+		filepath.Join(dir, "source"),
+		filepath.Join(dir, "missing-store"),
+		indexPath)
+
+	require.NoError(t,
+		os.WriteFile(configPath, []byte(contents), configFileMode))
+
+	// The PID lock lives under xdg.DataHome, which xdg resolves at
+	// package init; point it at the temp dir so the test neither
+	// touches nor collides with the real one.
+	t.Setenv("XDG_DATA_HOME", filepath.Join(dir, "data"))
+	xdg.Reload()
+	t.Cleanup(xdg.Reload)
+
+	return configPath, indexPath
+}
+
+// writeUnusableDestinationConfig builds a config whose destination
+// directory does not exist, which fails `remote info`, and whose local
+// index is bound to another destination, which fails `prune` and
+// `snapshot remove` (a missing destination alone only makes `snapshot
+// remove` warn). Returns the config path.
+func writeUnusableDestinationConfig(t *testing.T) string {
+	t.Helper()
+
+	configPath, indexPath := writeMissingDestinationConfig(t)
+
+	ctx := context.Background()
+
+	db, err := database.New(ctx, indexPath)
+	require.NoError(t, err)
+
+	defer func() { require.NoError(t, db.Close()) }()
+
+	require.NoError(t, database.NewRepositories(db).LocalMeta.Set(ctx,
+		database.LocalMetaKeyStorageURL, "file://"+t.TempDir()))
+
+	return configPath
+}
+
+// runEntry runs Entry with args after the program name and returns its
+// exit code and what it wrote to stdout and stderr.
+//
+// Not parallel-safe: it replaces os.Args, os.Stdout and os.Stderr.
+func runEntry(t *testing.T, args ...string) (int, string, string) {
+	t.Helper()
+
+	previousArgs := os.Args
+
+	t.Cleanup(func() {
+		os.Args = previousArgs
+		rootFlags = RootFlags{}
+	})
+
+	os.Args = append([]string{programName}, args...)
+
+	var code int
+
+	stdout, stderr := captureProcessStdoutAndStderr(t,
+		func() { code = Entry() })
+
+	return code, stdout, stderr
+}

internal/cli/entry_unknown_subcommand_test.go

@@ -0,0 +1,43 @@
+package cli //nolint:testpackage // shares runEntry and the command-word constants
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+// TestEntryGroupRejectsUnknownSubcommand runs each group command with a
+// mistyped subcommand, as a crontab line might carry one. It must fail
+// with exit 1 and name the word, as `vaultik bogus` does at the top
+// level, instead of printing the group's help and exiting 0 having done
+// nothing. The group name alone still prints its help.
+//
+//nolint:paralleltest // replaces os.Args, os.Stdout and os.Stderr
+func TestEntryGroupRejectsUnknownSubcommand(t *testing.T) {
+	for _, testCase := range []struct {
+		group    string
+		mistyped string
+	}{
+		{group: cmdSnapshot, mistyped: "craete"},
+		{group: "config", mistyped: "inti"},
+		{group: cmdRemote, mistyped: "ifno"},
+		{group: "database", mistyped: "delte"},
+		{group: "completion", mistyped: "nosuchshell"},
+	} {
+		t.Run(testCase.group, func(t *testing.T) {
+			code, stdout, stderr := runEntry(t,
+				testCase.group, testCase.mistyped)
+
+			assert.Equal(t, 1, code)
+			assert.Empty(t, stdout)
+			assert.Contains(t, stderr,
+				`unknown command "`+testCase.mistyped+`"`)
+
+			code, stdout, _ = runEntry(t, testCase.group)
+
+			assert.Equal(t, 0, code)
+			assert.Contains(t, stdout,
+				programName+" "+testCase.group+" [command]")
+		})
+	}
+}

internal/cli/info.go

@@ -1,12 +1,7 @@
 package cli
 
 import (
-	"context"
-	"errors"
-	"os"
-
 	"github.com/spf13/cobra"
-	"go.uber.org/fx"
 	"sneak.berlin/go/vaultik/internal/log"
 	"sneak.berlin/go/vaultik/internal/vaultik"
 )
@@ -33,44 +28,19 @@ func NewInfoCommand() *cobra.Command {
 			// Use the app framework
 			rootFlags := GetRootFlags()
 
-			return RunWithApp(cmd.Context(), AppOptions{
+			return RunOperation(cmd.Context(), AppOptions{
 				ConfigPath: configPath,
 				LogOptions: log.Options{
 					Verbose: rootFlags.Verbose,
 					Debug:   rootFlags.Debug,
 					Quiet:   rootFlags.Quiet,
 				},
-				Modules: []fx.Option{},
-				Invokes: []fx.Option{
-					fx.Invoke(func(v *vaultik.Vaultik, lc fx.Lifecycle) {
-						lc.Append(fx.Hook{
-							OnStart: func(_ context.Context) error {
-								go func() {
-									err := v.ShowInfo()
-									if err != nil {
-										if !errors.Is(err, context.Canceled) {
+				Mode: readOnly,
+			}, func(v *vaultik.Vaultik) error {
+				return v.ShowInfo()
+			}, func(err error) {
 				log.Error("Failed to show info", "error", err)
 				ReportErrorf("Failed to show info: %v", err)
-											os.Exit(1)
-										}
-									}
-
-									err = v.Shutdowner.Shutdown()
-									if err != nil {
-										log.Error("Failed to shutdown", "error", err)
-									}
-								}()
-
-								return nil
-							},
-							OnStop: func(_ context.Context) error {
-								v.Cancel()
-
-								return nil
-							},
-						})
-					}),
-				},
 			})
 		},
 	}

internal/cli/json_help_test.go

@@ -0,0 +1,24 @@
+package cli //nolint:testpackage // exercises the unexported command constructor
+
+import (
+	"testing"
+
+	"github.com/spf13/cobra"
+	"github.com/stretchr/testify/assert"
+)
+
+// TestJSONHelpSaysConfirmationPromptIsSkipped checks that the --json help
+// of `snapshot remove` and `prune` says the flag skips the confirmation
+// prompt. Both delete without asking under --json, since a prompt on
+// stdout would break the JSON document.
+func TestJSONHelpSaysConfirmationPromptIsSkipped(t *testing.T) {
+	t.Parallel()
+
+	for _, cmd := range []*cobra.Command{
+		newSnapshotRemoveCommand(),
+		NewPruneCommand(),
+	} {
+		assert.Contains(t, cmd.Flags().Lookup("json").Usage,
+			"skips the confirmation prompt", cmd.Name())
+	}
+}

internal/cli/prune.go

@@ -1,12 +1,7 @@
 package cli
 
 import (
-	"context"
-	"errors"
-	"os"
-
 	"github.com/spf13/cobra"
-	"go.uber.org/fx"
 	"sneak.berlin/go/vaultik/internal/log"
 	"sneak.berlin/go/vaultik/internal/vaultik"
 )
@@ -18,18 +13,34 @@ func NewPruneCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:   "prune",
 		Short: "Tidy local database and remote storage",
-		Long: `Removes orphaned data from both the local index database and
-unreferenced blobs from the backup destination store.
+		Long: `Removes orphaned data from the local index database, and
+unreferenced blobs and directories under metadata/ with no manifest
+from the backup destination store.
 
-Local cleanup drops incomplete snapshots and any files, chunks, or
+Needs permission to list, read and delete objects on the destination
+store. It does not need the age private key.
+
+Local cleanup drops snapshots whose manifest is gone from the
+destination store, incomplete snapshots, and any files, chunks, or
 blobs no longer referenced by a completed snapshot. Remote cleanup
 scans every snapshot manifest in the destination store, builds the
 set of still-referenced blob hashes, and deletes any blob not in that
-set.
+set that is older than prune_grace_period (default 24h). A newer one
+is kept: it may belong to a backup still running, on this host or
+another sharing the destination store, that has not yet uploaded its
+manifest. The grace period does not protect a blob such a backup
+reuses from an earlier run that stopped before its manifest: that
+blob keeps the time of its first upload, so it can be deleted.
+Remote cleanup also deletes each directory under metadata/ with no
+manifest, as a backup that stopped before its manifest leaves, once
+all its objects are older than prune_grace_period and its snapshot is
+not in the local index.
 
-Snapshot create --prune and snapshot remove run the same cleanup
-automatically; this command is the manual entry point for the same
-work (e.g. after a crashed backup or to reclaim storage).`,
+Snapshot create --prune runs the same cleanup automatically; this
+command is the manual entry point for the same work (e.g. after a
+crashed backup or to reclaim storage). Snapshot remove leaves blobs in
+place; run this command afterwards to delete the ones no longer
+referenced.`,
 		Args: cobra.NoArgs,
 		RunE: func(cmd *cobra.Command, _ []string) error {
 			// Use unified config resolution
@@ -41,57 +52,27 @@ work (e.g. after a crashed backup or to reclaim storage).`,
 			// Use the app framework like other commands
 			rootFlags := GetRootFlags()
 
-			return RunWithApp(cmd.Context(), AppOptions{
+			return RunOperation(cmd.Context(), AppOptions{
 				ConfigPath: configPath,
 				LogOptions: log.Options{
 					Verbose: rootFlags.Verbose,
 					Debug:   rootFlags.Debug,
-					Quiet:   rootFlags.Quiet || opts.JSON,
+					Quiet:   rootFlags.Quiet,
+					JSON:    opts.JSON,
 				},
-				Modules: []fx.Option{},
-				Invokes: []fx.Option{
-					fx.Invoke(func(v *vaultik.Vaultik, lc fx.Lifecycle) {
-						lc.Append(fx.Hook{
-							OnStart: func(_ context.Context) error {
-								// Start the prune operation in a goroutine
-								go func() {
-									// Run the prune operation
-									err := v.Prune(opts)
-									if err != nil {
-										if !errors.Is(err, context.Canceled) {
-											if !opts.JSON {
+				Mode: mutating,
+			}, func(v *vaultik.Vaultik) error {
+				return v.Prune(opts)
+			}, func(err error) {
 				log.Error("Prune operation failed", "error", err)
 				ReportErrorf("Prune failed: %v", err)
-											}
-
-											os.Exit(1)
-										}
-									}
-
-									// Shutdown the app when prune completes
-									err = v.Shutdowner.Shutdown()
-									if err != nil {
-										log.Error("Failed to shutdown", "error", err)
-									}
-								}()
-
-								return nil
-							},
-							OnStop: func(_ context.Context) error {
-								log.Debug("Stopping prune operation")
-								v.Cancel()
-
-								return nil
-							},
-						})
-					}),
-				},
 			})
 		},
 	}
 
 	cmd.Flags().BoolVar(&opts.Force, "force", false, "Skip confirmation prompt")
-	cmd.Flags().BoolVar(&opts.JSON, "json", false, "Output pruning stats as JSON")
+	cmd.Flags().BoolVar(&opts.JSON, "json", false,
+		"Output pruning stats as JSON; skips the confirmation prompt, as --force does")
 
 	return cmd
 }

internal/cli/quiet_test.go

@@ -0,0 +1,206 @@
+package cli //nolint:testpackage // sets the unexported rootFlags directly
+
+import (
+	"bytes"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/spf13/cobra"
+)
+
+// setRootFlags overrides the global rootFlags for the duration of one
+// test and restores it afterward. These tests must not run in parallel:
+// the flags are process-global, so the whole struct is saved and put
+// back rather than left mutated for the next test.
+func setRootFlags(t *testing.T, f RootFlags) {
+	t.Helper()
+
+	old := rootFlags
+	rootFlags = f
+
+	t.Cleanup(func() { rootFlags = old })
+}
+
+// seedFile writes content to a fresh file and returns its path.
+func seedFile(t *testing.T, dir, name, content string) string {
+	t.Helper()
+
+	path := filepath.Join(dir, name)
+
+	err := os.WriteFile(path, []byte(content), 0o600)
+	if err != nil {
+		t.Fatalf("seeding %s: %v", name, err)
+	}
+
+	return path
+}
+
+// mustExecute runs a command with its output captured and fails the test
+// if it errors, returning what the command printed.
+func mustExecute(t *testing.T, cmd *cobra.Command, args ...string) string {
+	t.Helper()
+
+	var out bytes.Buffer
+
+	cmd.SetOut(&out)
+	cmd.SetArgs(args)
+
+	err := cmd.Execute()
+	if err != nil {
+		t.Fatalf("%s failed: %v", cmd.Name(), err)
+	}
+
+	return out.String()
+}
+
+// TestVersionQuietSuppressesReport checks that --quiet silences the whole
+// version report: it is human-facing output, not a scriptable value.
+//
+//nolint:paralleltest // mutates the process-global rootFlags
+func TestVersionQuietSuppressesReport(t *testing.T) {
+	setRootFlags(t, RootFlags{Quiet: true})
+
+	out := mustExecute(t, NewVersionCommand())
+
+	if out != "" {
+		t.Errorf("--quiet version printed %q, want nothing", out)
+	}
+}
+
+// TestConfigGetIgnoresQuiet checks that a config value is printed even
+// under --quiet: it is scriptable output a caller depends on, so --quiet
+// must not suppress it, and it stays machine-plain (no marker, no color).
+//
+//nolint:paralleltest // mutates the process-global rootFlags
+func TestConfigGetIgnoresQuiet(t *testing.T) {
+	dir := t.TempDir()
+	path := seedFile(t, dir, "config.yml", "storage_url: file:///mnt/x\n")
+
+	setRootFlags(t, RootFlags{Quiet: true, ConfigPath: path})
+
+	out := mustExecute(t, newConfigGetCommand(), "storage_url")
+
+	if out != "file:///mnt/x\n" {
+		t.Errorf("config get --quiet = %q, want the plain value", out)
+	}
+}
+
+// TestConfigSetQuietSuppressesConfirmation checks that --quiet silences
+// the confirmation line while still writing the value to the file.
+//
+//nolint:paralleltest // mutates the process-global rootFlags
+func TestConfigSetQuietSuppressesConfirmation(t *testing.T) {
+	dir := t.TempDir()
+	path := seedFile(t, dir, "config.yml", "compression_level: 3\n")
+
+	setRootFlags(t, RootFlags{Quiet: true, ConfigPath: path})
+
+	out := mustExecute(t, newConfigSetCommand(), "compression_level", "9")
+
+	if out != "" {
+		t.Errorf("--quiet config set printed %q, want nothing", out)
+	}
+
+	data, err := os.ReadFile(path) //nolint:gosec // G304: test-controlled path
+	if err != nil {
+		t.Fatalf("reading config back: %v", err)
+	}
+
+	if !strings.Contains(string(data), "compression_level: 9") {
+		t.Errorf("config set did not write the value under --quiet:\n%s", data)
+	}
+}
+
+// TestConfigSetConfirmsWhenNotQuiet checks that the confirmation names
+// the key (styled) when --quiet is not set.
+//
+//nolint:paralleltest // mutates the process-global rootFlags
+func TestConfigSetConfirmsWhenNotQuiet(t *testing.T) {
+	dir := t.TempDir()
+	path := seedFile(t, dir, "config.yml", "compression_level: 3\n")
+
+	setRootFlags(t, RootFlags{ConfigPath: path})
+
+	out := mustExecute(t, newConfigSetCommand(), "compression_level", "9")
+
+	if !strings.Contains(out, "compression_level") {
+		t.Errorf("config set did not confirm the key: %q", out)
+	}
+}
+
+// TestConfigInitQuietSuppressesConfirmation checks that --quiet silences
+// the "config written" confirmation while still writing the file.
+//
+//nolint:paralleltest // mutates the process-global rootFlags
+func TestConfigInitQuietSuppressesConfirmation(t *testing.T) {
+	dir := t.TempDir()
+	path := filepath.Join(dir, "new-config.yml")
+
+	setRootFlags(t, RootFlags{Quiet: true, ConfigPath: path})
+
+	out := mustExecute(t, newConfigInitCommand())
+
+	if out != "" {
+		t.Errorf("--quiet config init printed %q, want nothing", out)
+	}
+
+	_, err := os.Stat(path)
+	if err != nil {
+		t.Errorf("config init did not write the file under --quiet: %v", err)
+	}
+}
+
+// seedDatabaseDeleteConfig writes a valid config whose index_path is a
+// seeded database file, and returns both paths.
+func seedDatabaseDeleteConfig(t *testing.T, dir string) (string, string) {
+	t.Helper()
+
+	dbPath := seedFile(t, dir, "index.sqlite", "not-a-real-db")
+	cfg := fmt.Sprintf(hermeticConfig,
+		filepath.Join(dir, "source"), filepath.Join(dir, "store"), dbPath)
+	cfgPath := seedFile(t, dir, "config.yml", cfg)
+
+	return dbPath, cfgPath
+}
+
+// TestDatabaseDeleteQuietSuppressesMessage checks that --quiet silences
+// the "database deleted" line while still removing the file.
+//
+//nolint:paralleltest // mutates the process-global rootFlags
+func TestDatabaseDeleteQuietSuppressesMessage(t *testing.T) {
+	dir := t.TempDir()
+	dbPath, cfgPath := seedDatabaseDeleteConfig(t, dir)
+
+	setRootFlags(t, RootFlags{Quiet: true, ConfigPath: cfgPath})
+
+	out := mustExecute(t, newDatabaseDeleteCommand(), "--force")
+
+	if out != "" {
+		t.Errorf("--quiet database delete printed %q, want nothing", out)
+	}
+
+	_, err := os.Stat(dbPath)
+	if !os.IsNotExist(err) {
+		t.Errorf("database delete did not remove the file: stat err = %v", err)
+	}
+}
+
+// TestDatabaseDeleteReportsWhenNotQuiet checks that the deletion is
+// reported when --quiet is not set.
+//
+//nolint:paralleltest // mutates the process-global rootFlags
+func TestDatabaseDeleteReportsWhenNotQuiet(t *testing.T) {
+	dir := t.TempDir()
+	_, cfgPath := seedDatabaseDeleteConfig(t, dir)
+
+	setRootFlags(t, RootFlags{ConfigPath: cfgPath})
+
+	out := mustExecute(t, newDatabaseDeleteCommand(), "--force")
+
+	if !strings.Contains(out, "deleted") {
+		t.Errorf("database delete did not report the deletion: %q", out)
+	}
+}

internal/cli/remote.go

@@ -1,12 +1,9 @@
 package cli
 
 import (
-	"context"
 	"errors"
-	"os"
 
 	"github.com/spf13/cobra"
-	"go.uber.org/fx"
 	"sneak.berlin/go/vaultik/internal/log"
 	"sneak.berlin/go/vaultik/internal/vaultik"
 )
@@ -21,6 +18,12 @@ func NewRemoteCommand() *cobra.Command {
 		Use:   "remote",
 		Short: "Remote storage management commands",
 		Long:  "Commands for inspecting and managing remote storage",
+		// cobra checks Args only on a command that has a Run: without
+		// one, a mistyped subcommand printed help and exited 0.
+		Args: cobra.NoArgs,
+		Run: func(cmd *cobra.Command, _ []string) {
+			_ = cmd.Help()
+		},
 	}
 
 	// Add subcommands
@@ -37,18 +40,20 @@ func newRemoteNukeCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:   "nuke",
 		Short: "Delete ALL snapshot metadata and blobs from the backup destination store",
-		Long: `Removes every snapshot's metadata and every blob from remote
-storage. After this command completes successfully the bucket prefix is
-empty and the next backup starts from scratch.
+		Long: `Removes every snapshot's metadata, every blob and the destination
+marker from the backup destination store, and each removed snapshot from
+the local index. After this command completes successfully the bucket
+prefix is empty and the next backup starts from scratch.
 
-This is destructive and irreversible. Requires --force.`,
+This is destructive and irreversible. Requires --force, and permission
+to list and delete objects on the destination store.`,
 		Args: cobra.NoArgs,
 		RunE: func(cmd *cobra.Command, _ []string) error {
 			if !force {
 				return errNukeNeedsForce
 			}
 
-			return runVaultikApp(cmd, false, false, "Remote nuke failed",
+			return runVaultikApp(cmd, mutating, false, "Remote nuke failed",
 				func(v *vaultik.Vaultik) error {
 					return v.NukeRemote(true)
 				})
@@ -72,7 +77,10 @@ func newRemoteInfoCommand() *cobra.Command {
 - Size of all snapshot metadata (per snapshot and total)
 - Count and total size of all blobs
 - Count and size of referenced blobs (from all manifests)
-- Count and size of orphaned blobs (not referenced by any manifest)`,
+- Count and size of orphaned blobs (not referenced by any manifest)
+
+Needs permission to list and read the destination store. It does not
+need the age private key.`,
 		Args: cobra.NoArgs,
 		RunE: func(cmd *cobra.Command, _ []string) error {
 			// Use unified config resolution
@@ -83,47 +91,20 @@ func newRemoteInfoCommand() *cobra.Command {
 
 			rootFlags := GetRootFlags()
 
-			return RunWithApp(cmd.Context(), AppOptions{
+			return RunOperation(cmd.Context(), AppOptions{
 				ConfigPath: configPath,
 				LogOptions: log.Options{
 					Verbose: rootFlags.Verbose,
 					Debug:   rootFlags.Debug,
-					Quiet:   rootFlags.Quiet || jsonOutput,
+					Quiet:   rootFlags.Quiet,
+					JSON:    jsonOutput,
 				},
-				Modules: []fx.Option{},
-				Invokes: []fx.Option{
-					fx.Invoke(func(v *vaultik.Vaultik, lc fx.Lifecycle) {
-						lc.Append(fx.Hook{
-							OnStart: func(_ context.Context) error {
-								go func() {
-									err := v.RemoteInfo(jsonOutput)
-									if err != nil {
-										if !errors.Is(err, context.Canceled) {
-											if !jsonOutput {
+				Mode: readOnly,
+			}, func(v *vaultik.Vaultik) error {
+				return v.RemoteInfo(jsonOutput)
+			}, func(err error) {
 				log.Error("Failed to get remote info", "error", err)
 				ReportErrorf("Failed to get remote info: %v", err)
-											}
-
-											os.Exit(1)
-										}
-									}
-
-									err = v.Shutdowner.Shutdown()
-									if err != nil {
-										log.Error("Failed to shutdown", "error", err)
-									}
-								}()
-
-								return nil
-							},
-							OnStop: func(_ context.Context) error {
-								v.Cancel()
-
-								return nil
-							},
-						})
-					}),
-				},
 			})
 		},
 	}

internal/cli/root.go

@@ -9,6 +9,7 @@ import (
 
 	"github.com/adrg/xdg"
 	"github.com/spf13/cobra"
+	"sneak.berlin/go/vaultik/internal/ui"
 )
 
 // errConfigNotFound is wrapped by all config-resolution failures.
@@ -34,9 +35,10 @@ func NewRootCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:   "vaultik",
 		Short: "Secure incremental backup tool with asymmetric encryption",
-		Long: `vaultik is a secure incremental backup tool that encrypts data using age
-public keys and uploads to S3-compatible storage. No private keys are needed
-on the source system.`,
+		Long: `vaultik is an incremental backup tool that encrypts data to age public
+keys and uploads it to an S3-compatible object store, a local directory or
+an rclone remote. The backed-up machine needs no private key: only
+'snapshot restore' and 'snapshot verify --deep' need one.`,
 		SilenceUsage: true,
 		// Bare 'vaultik' (no subcommand): print help. The banner is
 		// printed once at process startup by Entry, before cobra
@@ -49,16 +51,20 @@ on the source system.`,
 
 	// Add global flags
 	cmd.PersistentFlags().StringVar(&rootFlags.ConfigPath, "config", "",
-		"Path to config file (default: $VAULTIK_CONFIG or platform config dir)")
+		"Path to config file (default: $VAULTIK_CONFIG, then the platform "+
+			"config directory, then /etc/vaultik/config.yml)")
 	cmd.PersistentFlags().BoolVarP(&rootFlags.Verbose, "verbose", "v", false,
 		"Enable verbose output")
 	cmd.PersistentFlags().BoolVar(&rootFlags.Debug, "debug", false,
 		"Enable debug output")
 	cmd.PersistentFlags().BoolVarP(&rootFlags.Quiet, "quiet", "q", false,
-		"Suppress non-error output")
+		"Suppress status output (warnings, errors, JSON documents, "+
+			"'config get' values and confirmation prompts are still printed)")
 	cmd.PersistentFlags().BoolVar(&rootFlags.SkipErrors, "skip-errors", false,
-		"Continue past per-file errors instead of aborting "+
-			"(applies to snapshot create and restore)")
+		"Skip files that cannot be read when creating a snapshot, or "+
+			"that cannot be restored when restoring, instead of aborting "+
+			"(packing and storage errors while creating a snapshot still "+
+			"abort)")
 
 	// Add subcommands
 	cmd.AddCommand(
@@ -71,6 +77,21 @@ on the source system.`,
 		NewDatabaseCommand(),
 	)
 
+	// cobra adds the completion command only when Execute runs, and with
+	// no Run, so it skipped its own cobra.NoArgs and a mistyped shell name
+	// printed help and exited 0. Adding it here lets it have a Run. Its
+	// scripts go to os.Stdout as it is now: a later SetOut does not
+	// redirect them.
+	cmd.InitDefaultCompletionCmd()
+
+	for _, sub := range cmd.Commands() {
+		if sub.Name() == "completion" {
+			sub.Run = func(completion *cobra.Command, _ []string) {
+				_ = completion.Help()
+			}
+		}
+	}
+
 	return cmd
 }
 
@@ -80,6 +101,20 @@ func GetRootFlags() RootFlags {
 	return rootFlags
 }
 
+// commandUI returns a UI writer for a command's stdout, in quiet mode
+// when the global --quiet flag is set. This is how the pure-cli
+// commands (version, config, database) reach internal/ui: color follows
+// the writer (a TTY gets color, a captured test buffer does not), and
+// --quiet silences the same message classes it silences everywhere else.
+func commandUI(cmd *cobra.Command) *ui.Writer {
+	w := ui.New(cmd.OutOrStdout())
+	if GetRootFlags().Quiet {
+		w.SetQuiet(true)
+	}
+
+	return w
+}
+
 // ResolveConfigPath resolves the config file path from flags, environment, or default.
 // Search order: --config flag, VAULTIK_CONFIG env, XDG config dir,
 // /etc/vaultik/config.yml.

internal/cli/runapp_interrupt_test.go

@@ -0,0 +1,97 @@
+package cli_test
+
+import (
+	"context"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+	"go.uber.org/fx"
+	"sneak.berlin/go/vaultik/internal/cli"
+)
+
+// TestRunAppWaitsForOperationCleanupOnShutdown drives RunApp with an fx app
+// wired the way RunOperation wires a command: a single lifecycle hook whose
+// OnStart launches the operation in its own goroutine and whose OnStop cancels
+// it and blocks until that goroutine returns. The operation stands in for a
+// restore blocked mid-download — it holds a decrypted "scratch" file and only
+// removes it as it unwinds on cancellation.
+//
+// The app is asked to stop once the operation is running (standing in for an
+// OS interrupt; fx delivers a real signal and Shutdowner.Shutdown() on the
+// same app.Wait channel, so both drive the identical shutdown path). RunApp
+// must not return until app.Stop has run the OnStop hook, so the scratch file
+// must be gone by the time RunApp returns. Before the fix RunApp returned as
+// soon as the app.Wait/Done channel fired, without running app.Stop, so the
+// cleanup never ran and this file would still be on disk (issue #159).
+func TestRunAppWaitsForOperationCleanupOnShutdown(t *testing.T) {
+	t.Parallel()
+
+	scratch := filepath.Join(t.TempDir(), "decrypted-scratch")
+	require.NoError(t, os.WriteFile(scratch, []byte("secret"), 0o600))
+
+	// Cancel and reap the operation even if RunApp returns without doing so
+	// (the buggy path), so the goroutine cannot leak past the test.
+	opCtx, opCancel := context.WithCancel(context.Background())
+	t.Cleanup(opCancel)
+
+	var stop func(context.Context) bool
+
+	app := fx.New(
+		fx.NopLogger,
+		fx.Invoke(func(lc fx.Lifecycle, sh fx.Shutdowner) {
+			lc.Append(fx.Hook{
+				OnStart: func(_ context.Context) error {
+					done := make(chan struct{})
+
+					go func() {
+						defer close(done)
+
+						// Blocked mid-operation until cancelled, then run the
+						// cleanup an interrupted restore would run.
+						<-opCtx.Done()
+
+						_ = os.Remove(scratch)
+					}()
+
+					stop = func(ctx context.Context) bool {
+						opCancel()
+
+						select {
+						case <-done:
+							return true
+						case <-ctx.Done():
+							return false
+						}
+					}
+
+					// Ask the app to stop now that the operation is running.
+					go func() { _ = sh.Shutdown() }()
+
+					return nil
+				},
+				OnStop: func(ctx context.Context) error {
+					stop(ctx)
+
+					return nil
+				},
+			})
+		}),
+	)
+
+	done := make(chan error, 1)
+	go func() { done <- cli.RunApp(context.Background(), app) }()
+
+	select {
+	case err := <-done:
+		require.NoError(t, err)
+	case <-time.After(30 * time.Second):
+		t.Fatal("RunApp did not return after shutdown was requested")
+	}
+
+	_, err := os.Stat(scratch)
+	require.True(t, os.IsNotExist(err),
+		"RunApp returned before the operation removed its decrypted scratch file")
+}

internal/cli/snapshot.go

@@ -1,13 +1,10 @@
 package cli
 
 import (
-	"context"
 	"errors"
 	"fmt"
-	"os"
 
 	"github.com/spf13/cobra"
-	"go.uber.org/fx"
 	"sneak.berlin/go/vaultik/internal/log"
 	"sneak.berlin/go/vaultik/internal/vaultik"
 )
@@ -44,6 +41,12 @@ func NewSnapshotCommand() *cobra.Command {
 		Use:   "snapshot",
 		Short: "Snapshot management commands",
 		Long:  "Commands for creating, listing, and managing snapshots",
+		// cobra checks Args only on a command that has a Run: without
+		// one, a mistyped subcommand printed help and exited 0.
+		Args: cobra.NoArgs,
+		Run: func(cmd *cobra.Command, _ []string) {
+			_ = cmd.Help()
+		},
 	}
 
 	// Add subcommands
@@ -57,6 +60,21 @@ func NewSnapshotCommand() *cobra.Command {
 	return cmd
 }
 
+// snapshotCreateLong is the long help text of 'snapshot create'.
+const snapshotCreateLong = `Creates new snapshots of the configured directories.
+
+If snapshot names are provided, only those snapshots are created.
+If no names are provided, all configured snapshots are created.
+
+Needs at least one public key in age_recipients, and permission to write
+objects at the backup destination store. It does not need the age private
+key. With --prune it also needs permission to list, read and delete objects
+there.
+
+The config is read from the path given by --config or VAULTIK_CONFIG;
+otherwise from the platform config directory (~/.config/vaultik/config.yml
+on Linux), then /etc/vaultik/config.yml.`
+
 // newSnapshotCreateCommand creates the 'snapshot create' subcommand
 func newSnapshotCreateCommand() *cobra.Command {
 	opts := &vaultik.SnapshotCreateOptions{}
@@ -64,15 +82,14 @@ func newSnapshotCreateCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:   "create [snapshot-names...]",
 		Short: "Create new snapshots",
-		Long: `Creates new snapshots of the configured directories.
-
-If snapshot names are provided, only those snapshots are created.
-If no names are provided, all configured snapshots are created.
-
-Config is located at /etc/vaultik/config.yml by default, but can be overridden by
-specifying a path using --config or by setting VAULTIK_CONFIG to a path.`,
+		Long:  snapshotCreateLong,
 		Args:  cobra.ArbitraryArgs,
 		RunE: func(cmd *cobra.Command, args []string) error {
+			err := opts.Validate()
+			if err != nil {
+				return err
+			}
+
 			// Pass snapshot names from args
 			opts.Snapshots = args
 			// --skip-errors is a global flag on the root command.
@@ -86,7 +103,8 @@ specifying a path using --config or by setting VAULTIK_CONFIG to a path.`,
 			// Use the backup functionality from cli package
 			rootFlags := GetRootFlags()
 
-			return RunWithApp(cmd.Context(), AppOptions{
+			// --cron suppression is wired through v.UI by setupGlobals.
+			return RunOperation(cmd.Context(), AppOptions{
 				ConfigPath: configPath,
 				LogOptions: log.Options{
 					Verbose: rootFlags.Verbose,
@@ -94,54 +112,46 @@ specifying a path using --config or by setting VAULTIK_CONFIG to a path.`,
 					Cron:    opts.Cron,
 					Quiet:   rootFlags.Quiet,
 				},
-				Modules: []fx.Option{},
-				Invokes: []fx.Option{
-					fx.Invoke(func(v *vaultik.Vaultik, lc fx.Lifecycle) {
-						lc.Append(fx.Hook{
-							OnStart: func(_ context.Context) error {
-								// Start the snapshot creation in a goroutine
-								go func() {
-									// --cron suppression is wired through v.UI by setupGlobals.
-									err := v.CreateSnapshot(opts)
-									if err != nil {
-										if !errors.Is(err, context.Canceled) {
+				Mode: mutating,
+			}, func(v *vaultik.Vaultik) error {
+				return v.CreateSnapshot(opts)
+			}, func(err error) {
+				if errors.Is(err, vaultik.ErrPostBackupPruneFailed) {
+					log.Error("Post-backup prune failed", "error", err)
+					ReportErrorf("Backup completed, but %v", err)
+
+					return
+				}
+
+				if errors.Is(err, vaultik.ErrFilesUnreadable) ||
+					errors.Is(err, vaultik.ErrSourcePathEmpty) ||
+					errors.Is(err, vaultik.ErrVolumeNotMounted) {
+					log.Error("Backup completed with an error", "error", err)
+					ReportErrorf("Backup completed, but %v", err)
+
+					return
+				}
+
 				log.Error("Snapshot creation failed", "error", err)
 				ReportErrorf("Snapshot creation failed: %v", err)
-											os.Exit(1)
-										}
-									}
-
-									// Shutdown the app when snapshot completes
-									err = v.Shutdowner.Shutdown()
-									if err != nil {
-										log.Error("Failed to shutdown", "error", err)
-									}
-								}()
-
-								return nil
-							},
-							OnStop: func(_ context.Context) error {
-								log.Debug("Stopping snapshot creation")
-								// Cancel the Vaultik context
-								v.Cancel()
-
-								return nil
-							},
-						})
-					}),
-				},
 			})
 		},
 	}
 
 	cmd.Flags().BoolVar(&opts.Cron, "cron", false,
-		"Run in cron mode (silent unless error)")
+		"Run in cron mode (silent unless warning or error)")
 	cmd.Flags().BoolVar(&opts.Prune, "prune", false,
-		"After backup, drop older snapshots of the same name and remove "+
-			"orphaned blobs")
+		"After backup, drop older snapshots of the same name, remove "+
+			"orphaned blobs older than prune_grace_period, and delete each "+
+			"directory under metadata/ with no manifest once all its "+
+			"objects are older than prune_grace_period and its snapshot is "+
+			"not in the local index (not after a backup that skipped "+
+			"unreadable files, found a configured path empty, or did not "+
+			"find a mount point an earlier snapshot found)")
 	cmd.Flags().StringVar(&opts.KeepNewerThan, "keep-newer-than", "",
 		"With --prune: keep snapshots newer than this duration "+
-			"(e.g. 4w, 30d, 6mo) instead of only the latest")
+			"(e.g. 30d, 4w, 6mo, 1y; m is minutes, mo is months) "+
+			"instead of only the latest")
 
 	return cmd
 }
@@ -154,10 +164,17 @@ func newSnapshotListCommand() *cobra.Command {
 		Use:     "list",
 		Aliases: []string{"ls"},
 		Short:   "List all snapshots",
-		Long:    "Lists all snapshots with their ID, timestamp, and compressed size",
+		Long: `Lists the snapshots in the local index and on the backup destination
+store, with their ID, timestamp, and sizes.
+
+Needs permission to list and read the destination store. It does not need
+the age private key. A snapshot found only on the destination store is
+listed by its remote key, because its ID is stored only in the local index
+and in the snapshot's encrypted database. If the destination store cannot
+be listed, the command warns and lists the local index alone.`,
 		Args: cobra.NoArgs,
 		RunE: func(cmd *cobra.Command, _ []string) error {
-			return runVaultikApp(cmd, false, false,
+			return runVaultikApp(cmd, readOnly, false,
 				"Failed to list snapshots",
 				func(v *vaultik.Vaultik) error {
 					return v.ListSnapshots(jsonOutput)
@@ -177,11 +194,19 @@ func newSnapshotPurgeCommand() *cobra.Command {
 	cmd := &cobra.Command{
 		Use:   "purge",
 		Short: "Purge old snapshots",
-		Long: `Removes snapshots based on age or count criteria.
+		Long: `Removes snapshots based on age or count criteria. Exactly one of
+--keep-latest and --older-than is required.
 
-Retention is per-snapshot-name: --keep-latest keeps the latest of each
-configured snapshot name, not the latest globally. Use --snapshot to
-restrict the operation to specific snapshot names.`,
+Retention is per-snapshot-name: --keep-latest keeps the latest snapshot of
+each name, not the latest globally. Use --snapshot to restrict the
+operation to specific snapshot names.
+
+Only snapshots in the local index are removed. Each one is removed from
+the local index and its metadata is deleted from the backup destination
+store. Blobs are not touched: run 'vaultik prune' afterwards.
+
+Needs permission to list and delete objects on the destination store. If
+the destination store cannot be listed, nothing is removed.`,
 		Args: cobra.NoArgs,
 		RunE: func(cmd *cobra.Command, _ []string) error {
 			// Validate flags
@@ -193,7 +218,7 @@ restrict the operation to specific snapshot names.`,
 				return errPurgeCriteriaBoth
 			}
 
-			return runVaultikApp(cmd, false, false,
+			return runVaultikApp(cmd, mutating, false,
 				"Failed to purge snapshots",
 				func(v *vaultik.Vaultik) error {
 					return v.PurgeSnapshotsWithOptions(opts)
@@ -204,7 +229,8 @@ restrict the operation to specific snapshot names.`,
 	cmd.Flags().BoolVar(&opts.KeepLatest, "keep-latest", false,
 		"Keep only the latest snapshot of each name")
 	cmd.Flags().StringVar(&opts.OlderThan, "older-than", "",
-		"Remove snapshots older than duration (e.g., 30d, 6m, 1y)")
+		"Remove snapshots older than duration "+
+			"(e.g. 30d, 4w, 6mo, 1y; m is minutes, mo is months)")
 	cmd.Flags().BoolVar(&opts.Force, "force", false, "Skip confirmation prompt")
 	cmd.Flags().StringArrayVar(&opts.Names, "snapshot", nil,
 		"Restrict to snapshots with these names (repeat for multiple)")
@@ -218,8 +244,30 @@ func newSnapshotVerifyCommand() *cobra.Command {
 
 	cmd := &cobra.Command{
 		Use:   "verify <snapshot-id>",
-		Short: "Verify snapshot integrity",
-		Long:  "Verifies that all blobs referenced in a snapshot exist",
+		Short: "Check a snapshot's blobs are present with the listed size",
+		Long: `Checks that every blob the snapshot's manifest lists is present
+in storage with the size the manifest records, and that the
+snapshot's encrypted database is present. It does not read blob
+contents and does not need the age private key.
+
+Needs permission to read the backup destination store, and to list it
+when a remote key is given.
+
+Use --deep to download, decrypt, and re-hash every blob to detect
+corruption -- integrity, not who wrote it.
+
+--deep requires the age private key, in the VAULTIK_AGE_SECRET_KEY
+environment variable or as age_secret_key in the config file; the
+environment variable wins. The variable may hold the whole age-keygen
+file, with its comments and all of its keys. Read the key from the
+file rather than typing it, so that it does not land in your shell
+history:
+
+  export VAULTIK_AGE_SECRET_KEY="$(cat vaultik_backup_private_key.txt)"
+
+The snapshot may be named by its ID or, on a host with no local
+index, by the remote key that 'snapshot list' prints for a
+remote-only snapshot (an unambiguous leading part is enough).`,
 		Args: requireSnapshotIDArg,
 		RunE: func(cmd *cobra.Command, args []string) error {
 			snapshotID := args[0]
@@ -232,52 +280,32 @@ func newSnapshotVerifyCommand() *cobra.Command {
 
 			rootFlags := GetRootFlags()
 
-			return RunWithApp(cmd.Context(), AppOptions{
+			return RunOperation(cmd.Context(), AppOptions{
 				ConfigPath: configPath,
 				LogOptions: log.Options{
 					Verbose: rootFlags.Verbose,
 					Debug:   rootFlags.Debug,
-					Quiet:   rootFlags.Quiet || opts.JSON,
+					Quiet:   rootFlags.Quiet,
+					JSON:    opts.JSON,
 				},
-				Modules: []fx.Option{},
-				Invokes: []fx.Option{
-					fx.Invoke(func(v *vaultik.Vaultik, lc fx.Lifecycle) {
-						lc.Append(fx.Hook{
-							OnStart: func(_ context.Context) error {
-								go func() {
-									err := v.VerifySnapshotWithOptions(snapshotID, opts)
-									if err != nil {
-										if !errors.Is(err, context.Canceled) {
-											if !opts.JSON {
+				Mode: readOnly,
+			}, func(v *vaultik.Vaultik) error {
+				return v.VerifySnapshotWithOptions(snapshotID, opts)
+			}, func(err error) {
+				if opts.JSON {
+					return
+				}
+
 				log.Error("Verification failed", "error", err)
 				ReportErrorf("Verification failed: %v", err)
-											}
-
-											os.Exit(1)
-										}
-									}
-
-									err = v.Shutdowner.Shutdown()
-									if err != nil {
-										log.Error("Failed to shutdown", "error", err)
-									}
-								}()
-
-								return nil
-							},
-							OnStop: func(_ context.Context) error {
-								v.Cancel()
-
-								return nil
-							},
-						})
-					}),
-				},
 			})
 		},
 	}
 
-	cmd.Flags().BoolVar(&opts.Deep, "deep", false, "Download and verify blob hashes")
+	cmd.Flags().BoolVar(&opts.Deep, "deep", false,
+		"Download, decrypt and re-hash every blob; requires the age private "+
+			"key in VAULTIK_AGE_SECRET_KEY, or as age_secret_key in the config "+
+			"file")
 	cmd.Flags().BoolVar(&opts.JSON, "json", false, "Output verification results as JSON")
 
 	return cmd
@@ -297,21 +325,31 @@ By default, this removes the snapshot from the local index database and
 strips the snapshot's metadata from the backup destination store. Blobs
 are NOT touched: deleting them requires reading every remaining remote
 manifest (the destination store may hold snapshots this host doesn't
-know about), which is what 'vaultik prune' does. On success the command
-prints the exact 'vaultik prune' invocation to run as a follow-up.
+know about), which is what 'vaultik prune' does. The command prints the
+'vaultik prune' invocation to run as a follow-up only when it removed the
+snapshot's metadata from the destination store, and never under --json.
+
+The argument must be a snapshot ID: a remote key is not accepted. An ID
+that names no snapshot in the local index or on the destination store is
+an error. Needs permission to list and delete objects on the destination
+store, unless --local-only is given.
 
 Use --local-only to skip the remote half (e.g. when you want to forget a
-snapshot locally without touching the destination store).
+snapshot locally without touching the destination store); the snapshot
+must then be in the local index.
 
-If the remote is unreachable, the local-database removal still completes
-and a warning is emitted; rerun 'vaultik prune' once the destination store
-is reachable to finish remote cleanup.
+If removing the snapshot's metadata from the destination store fails (for
+example because the remote is unreachable), the local-database removal
+still completes and a warning is emitted; run 'vaultik snapshot remove
+<snapshot-id>' again once the destination store is reachable to remove
+the snapshot's metadata from it ('vaultik prune' does not). For a
+snapshot that is only on the destination store, that failure is an error.
 
 To wipe the entire destination store and start over, use 'vaultik remote
 nuke --force' — it is the single supported entry point for that.`,
 		Args: requireSnapshotIDArg,
 		RunE: func(cmd *cobra.Command, args []string) error {
-			return runVaultikApp(cmd, opts.JSON, opts.JSON,
+			return runVaultikApp(cmd, mutating, opts.JSON,
 				"Failed to remove snapshot",
 				func(v *vaultik.Vaultik) error {
 					_, err := v.RemoveSnapshot(args[0], opts)
@@ -324,7 +362,8 @@ nuke --force' — it is the single supported entry point for that.`,
 	cmd.Flags().BoolVarP(&opts.Force, "force", "f", false, "Skip confirmation prompt")
 	cmd.Flags().BoolVar(&opts.DryRun, "dry-run", false,
 		"Show what would be removed without removing")
-	cmd.Flags().BoolVar(&opts.JSON, "json", false, "Output result as JSON")
+	cmd.Flags().BoolVar(&opts.JSON, "json", false,
+		"Output result as JSON; skips the confirmation prompt, as --force does")
 	cmd.Flags().BoolVar(&opts.LocalOnly, "local-only", false,
 		"Skip remote cleanup; only touch the local index")
 

internal/cli/snapshot_restore.go

@@ -1,16 +1,8 @@
 package cli
 
 import (
-	"context"
-	"errors"
-	"os"
-
 	"github.com/spf13/cobra"
-	"go.uber.org/fx"
-	"sneak.berlin/go/vaultik/internal/config"
-	"sneak.berlin/go/vaultik/internal/globals"
 	"sneak.berlin/go/vaultik/internal/log"
-	"sneak.berlin/go/vaultik/internal/storage"
 	"sneak.berlin/go/vaultik/internal/vaultik"
 )
 
@@ -25,15 +17,6 @@ type RestoreOptions struct {
 	Verify    bool     // Verify restored files after restore
 }
 
-// RestoreApp contains all dependencies needed for restore
-type RestoreApp struct {
-	Globals    *globals.Globals
-	Config     *config.Config
-	Storage    storage.Storer
-	Vaultik    *vaultik.Vaultik
-	Shutdowner fx.Shutdowner
-}
-
 // newSnapshotRestoreCommand creates the 'snapshot restore' subcommand
 func newSnapshotRestoreCommand() *cobra.Command {
 	opts := &RestoreOptions{}
@@ -47,9 +30,27 @@ This command will restore files from the specified snapshot to the
 target directory.
 If no paths are specified, all files are restored.
 If paths are specified, only matching files/directories are restored.
+A path that matches nothing in the snapshot is named in a warning, and
+the command exits non-zero after restoring the paths that match.
 
-Requires the VAULTIK_AGE_SECRET_KEY environment variable to be set with
-the age private key.
+The snapshot may be named by its ID or, when restoring on a host with no
+local index, by the remote key that 'snapshot list' prints for a
+remote-only snapshot (an unambiguous leading part is enough).
+
+Each file is restored under its original absolute path inside the target
+directory. File owners are restored only when the command runs as root.
+
+Requires the age private key, in the VAULTIK_AGE_SECRET_KEY environment
+variable or as age_secret_key in the config file; the environment
+variable wins. The variable may hold the whole age-keygen file, with its
+comments and all of its keys. Read the key from the file rather than
+typing it, so that it does not land in your shell history:
+
+  export VAULTIK_AGE_SECRET_KEY="$(cat vaultik_backup_private_key.txt)"
+
+Also needs permission to read the backup destination store (and to list
+it when a remote key is given), and free space in the temporary directory
+for the decrypted blobs.
 
 Examples:
   # Restore entire snapshot
@@ -77,7 +78,8 @@ Examples:
 	return cmd
 }
 
-// runRestore parses arguments and runs the restore operation through the app framework
+// runRestore parses arguments and runs the restore operation through the
+// app framework.
 func runRestore(cmd *cobra.Command, args []string, opts *RestoreOptions) error {
 	snapshotID := args[0]
 
@@ -86,87 +88,31 @@ func runRestore(cmd *cobra.Command, args []string, opts *RestoreOptions) error {
 		opts.Paths = args[restoreMinArgs:]
 	}
 
-	// Use unified config resolution
 	configPath, err := ResolveConfigPath()
 	if err != nil {
 		return err
 	}
 
-	// Use the app framework like other commands
 	rootFlags := GetRootFlags()
 
-	return RunWithApp(cmd.Context(), AppOptions{
+	return RunOperation(cmd.Context(), AppOptions{
 		ConfigPath: configPath,
 		LogOptions: log.Options{
 			Verbose: rootFlags.Verbose,
 			Debug:   rootFlags.Debug,
 			Quiet:   rootFlags.Quiet,
 		},
-		Modules: buildRestoreModules(),
-		Invokes: buildRestoreInvokes(snapshotID, opts),
-	})
-}
-
-// buildRestoreModules returns the fx.Options for dependency injection in restore
-func buildRestoreModules() []fx.Option {
-	return []fx.Option{
-		fx.Provide(fx.Annotate(
-			func(g *globals.Globals, cfg *config.Config,
-				storer storage.Storer, v *vaultik.Vaultik, shutdowner fx.Shutdowner) *RestoreApp {
-				return &RestoreApp{
-					Globals:    g,
-					Config:     cfg,
-					Storage:    storer,
-					Vaultik:    v,
-					Shutdowner: shutdowner,
-				}
-			},
-		)),
-	}
-}
-
-// buildRestoreInvokes returns the fx.Options that wire up the restore lifecycle
-func buildRestoreInvokes(snapshotID string, opts *RestoreOptions) []fx.Option {
-	return []fx.Option{
-		fx.Invoke(func(app *RestoreApp, lc fx.Lifecycle) {
-			lc.Append(fx.Hook{
-				OnStart: func(_ context.Context) error {
-					// Start the restore operation in a goroutine
-					go func() {
-						// Run the restore operation
-						restoreOpts := &vaultik.RestoreOptions{
+		Mode: readOnly,
+	}, func(v *vaultik.Vaultik) error {
+		return v.Restore(&vaultik.RestoreOptions{
 			SnapshotID: snapshotID,
 			TargetDir:  opts.TargetDir,
 			Paths:      opts.Paths,
 			Verify:     opts.Verify,
-							SkipErrors: GetRootFlags().SkipErrors,
-						}
-
-						err := app.Vaultik.Restore(restoreOpts)
-						if err != nil {
-							if !errors.Is(err, context.Canceled) {
+			SkipErrors: rootFlags.SkipErrors,
+		})
+	}, func(err error) {
 		log.Error("Restore operation failed", "error", err)
 		ReportErrorf("Restore failed: %v", err)
-								os.Exit(1)
-							}
-						}
-
-						// Shutdown the app when restore completes
-						err = app.Shutdowner.Shutdown()
-						if err != nil {
-							log.Error("Failed to shutdown", "error", err)
-						}
-					}()
-
-					return nil
-				},
-				OnStop: func(_ context.Context) error {
-					log.Debug("Stopping restore operation")
-					app.Vaultik.Cancel()
-
-					return nil
-				},
 	})
-		}),
-	}
 }

internal/cli/snapshot_restore_test.go

@@ -0,0 +1,37 @@
+package cli //nolint:testpackage // exercises the unexported command constructor
+
+import (
+	"strings"
+	"testing"
+
+	"github.com/spf13/pflag"
+)
+
+// TestRestoreCommandDoesNotTakeKeyAsArgument guards the fix for the age
+// key being echoed on the command line: restore must take the key only
+// from the environment, never as a flag value, and its help must show the
+// file-based form rather than a literal key that would land in shell
+// history.
+func TestRestoreCommandDoesNotTakeKeyAsArgument(t *testing.T) {
+	t.Parallel()
+
+	cmd := newSnapshotRestoreCommand()
+
+	cmd.Flags().VisitAll(func(f *pflag.Flag) {
+		lower := strings.ToLower(f.Name)
+		for _, banned := range []string{"key", "secret", "age", "identity"} {
+			if strings.Contains(lower, banned) {
+				t.Errorf("restore must not accept the key as a flag; found --%s", f.Name)
+			}
+		}
+	})
+
+	help := cmd.Long
+	if strings.Contains(help, "AGE-SECRET-KEY-") {
+		t.Error("restore help must not show a literal age private key to type")
+	}
+
+	if !strings.Contains(help, "$(cat ") {
+		t.Error("restore help should read the key from a file, e.g. $(cat ...)")
+	}
+}

internal/cli/vaultik_snapshot_types.go

@@ -1,12 +0,0 @@
-package cli
-
-import "time"
-
-// SnapshotInfo represents snapshot information for listing
-//
-//nolint:tagliatelle // snake_case is the established output format
-type SnapshotInfo struct {
-	ID             string    `json:"id"`
-	Timestamp      time.Time `json:"timestamp"`
-	CompressedSize int64     `json:"compressed_size"`
-}

internal/cli/version.go

@@ -2,11 +2,11 @@ package cli
 
 import (
 	"fmt"
-	"os"
 	"runtime"
 
 	"github.com/spf13/cobra"
 	"sneak.berlin/go/vaultik/internal/globals"
+	"sneak.berlin/go/vaultik/internal/ui"
 )
 
 // NewVersionCommand creates the version command
@@ -16,28 +16,44 @@ func NewVersionCommand() *cobra.Command {
 		Short: "Print version information",
 		Long:  `Print version, git commit, and build information for vaultik.`,
 		Args:  cobra.NoArgs,
-		Run: func(_ *cobra.Command, _ []string) {
-			_, _ = fmt.Fprintf(os.Stdout, "vaultik %s\n", globals.Version)
-			_, _ = fmt.Fprintf(os.Stdout, "  commit:     %s\n", globals.Commit)
-			_, _ = fmt.Fprintf(os.Stdout, "  build date: %s\n", globals.CommitDate)
-			_, _ = fmt.Fprintf(os.Stdout, "  go:         %s\n", runtime.Version())
-			_, _ = fmt.Fprintf(os.Stdout, "  os/arch:    %s/%s\n",
-				runtime.GOOS, runtime.GOARCH)
-			_, _ = fmt.Fprintf(os.Stdout, "  author:     %s\n", globals.Author)
-			_, _ = fmt.Fprintf(os.Stdout, "  homepage:   %s\n", globals.Homepage)
-			_, _ = fmt.Fprintf(os.Stdout, "  license:    %s\n", globals.License)
-
-			if globals.Version == "dev" {
-				_, _ = fmt.Fprintln(os.Stdout)
-				_, _ = fmt.Fprintln(os.Stdout,
-					"This is a development build (no version information embedded).")
-				_, _ = fmt.Fprintln(os.Stdout,
-					"Build a release binary with 'make vaultik' or download from")
-				_, _ = fmt.Fprintln(os.Stdout,
-					"https://sneak.berlin/go/vaultik for embedded version metadata.")
-			}
+		Run: func(cmd *cobra.Command, _ []string) {
+			writeVersion(commandUI(cmd))
 		},
 	}
 
 	return cmd
 }
+
+// writeVersion prints the version report through the UI writer. The
+// report is the output this command exists to produce, so it is written
+// plain (markers would corrupt the aligned report) via the writer's
+// underlying stdout; --quiet silences it like any other non-error
+// output.
+func writeVersion(out *ui.Writer) {
+	if out.Quiet() {
+		return
+	}
+
+	w := out.Out()
+
+	_, _ = fmt.Fprintf(w, "vaultik %s\n", globals.Version)
+	_, _ = fmt.Fprintf(w, "  commit:     %s\n", globals.Commit)
+	_, _ = fmt.Fprintf(w, "  build date: %s\n", globals.CommitDate)
+	_, _ = fmt.Fprintf(w, "  go:         %s\n", runtime.Version())
+	_, _ = fmt.Fprintf(w, "  os/arch:    %s/%s\n", runtime.GOOS, runtime.GOARCH)
+	_, _ = fmt.Fprintf(w, "  author:     %s\n", globals.Author)
+	_, _ = fmt.Fprintf(w, "  homepage:   %s\n", globals.Homepage)
+	_, _ = fmt.Fprintf(w, "  license:    %s\n", globals.License)
+
+	if globals.IsDevVersion(globals.Version) {
+		_, _ = fmt.Fprintln(w)
+		_, _ = fmt.Fprintln(w,
+			"This is a development build: it was not built from a tagged")
+		_, _ = fmt.Fprintln(w,
+			"commit, so it carries no release version. Released binaries")
+		_, _ = fmt.Fprintf(w,
+			"are published at %s\n", globals.ReleasesURL)
+		_, _ = fmt.Fprintln(w,
+			"and report their tag on the first line above.")
+	}
+}

internal/cli/version_test.go

@@ -0,0 +1,77 @@
+package cli_test
+
+import (
+	"bytes"
+	"strings"
+	"testing"
+
+	"sneak.berlin/go/vaultik/internal/cli"
+	"sneak.berlin/go/vaultik/internal/globals"
+)
+
+// runVersionCommand executes `vaultik version` with its output
+// captured, and returns what it printed.
+func runVersionCommand(t *testing.T) string {
+	t.Helper()
+
+	cmd := cli.NewVersionCommand()
+
+	var out bytes.Buffer
+
+	cmd.SetOut(&out)
+	cmd.SetErr(&out)
+	cmd.SetArgs([]string{})
+
+	err := cmd.Execute()
+	if err != nil {
+		t.Fatalf("version command failed: %v", err)
+	}
+
+	return out.String()
+}
+
+// TestVersionCommandReportsBuildVersion checks that the first line of
+// the report is the version the binary was actually built with. The
+// test binary carries no -ldflags, so that is the "dev" default -- the
+// same string an untagged `make vaultik` build stamps a prefix of.
+//
+//nolint:paralleltest // executes a command that reads the global rootFlags
+func TestVersionCommandReportsBuildVersion(t *testing.T) {
+	out := runVersionCommand(t)
+
+	wantFirst := "vaultik " + globals.Version
+	if first, _, _ := strings.Cut(out, "\n"); first != wantFirst {
+		t.Errorf("first line = %q, want %q", first, wantFirst)
+	}
+
+	if !strings.Contains(out, "commit:") {
+		t.Error("output does not report the commit")
+	}
+}
+
+// TestVersionCommandFlagsDevelopmentBuild is the regression test for
+// the thing this command exists to prevent: a build that is not a
+// release must say so. The notice used to be gated on the version
+// being exactly "dev", so once untagged builds started carrying their
+// commit sha it would have gone silent and an unreleased binary would
+// have looked like a release.
+//
+//nolint:paralleltest // executes a command that reads the global rootFlags
+func TestVersionCommandFlagsDevelopmentBuild(t *testing.T) {
+	if !globals.IsDevVersion(globals.Version) {
+		t.Skipf("test binary was stamped with release version %q",
+			globals.Version)
+	}
+
+	out := runVersionCommand(t)
+
+	if !strings.Contains(out, "development build") {
+		t.Errorf("dev build did not print the development-build notice:\n%s",
+			out)
+	}
+
+	if !strings.Contains(out, globals.ReleasesURL) {
+		t.Errorf("development-build notice does not point at %s:\n%s",
+			globals.ReleasesURL, out)
+	}
+}

internal/config/config.go

@@ -4,47 +4,73 @@
 package config
 
 import (
+	"bytes"
 	"errors"
 	"fmt"
 	"os"
 	"path/filepath"
 	"sort"
 	"strings"
+	"time"
 
 	"filippo.io/age"
 	"git.eeqj.de/sneak/smartconfig"
 	"github.com/adrg/xdg"
 	"go.uber.org/fx"
 	"gopkg.in/yaml.v3"
+	"sneak.berlin/go/vaultik/internal/chunker"
 	"sneak.berlin/go/vaultik/internal/log"
 )
 
 const appName = "vaultik"
 
-// Defaults and validation bounds for tunable settings.
+// secretKeyPrefix marks an age secret (private) key. It is compared
+// case-insensitively so a recipient entry that is actually a private key is
+// caught and never passed to age or echoed back.
+//
+//nolint:gosec // G101: marker for detecting a pasted secret key, not a credential
+const secretKeyPrefix = "AGE-SECRET-KEY-"
+
+// Defaults and validation bounds for tunable settings. MB and GB are
+// 1000-based, as the size parser reads them; the S3 part sizes are S3's own
+// limits, which are 1024-based (MiB, GiB).
 const (
-	defaultBlobSizeLimit    = Size(10 * 1024 * 1024 * 1024) // 10GB
-	defaultChunkSize        = Size(10 * 1024 * 1024)        // 10MB
-	defaultS3PartSize       = Size(5 * 1024 * 1024)         // 5MB
+	defaultBlobSizeLimit    = Size(10 * 1000 * 1000 * 1000) // 10GB
+	defaultChunkSize        = Size(10 * 1000 * 1000)        // 10MB
+	defaultS3PartSize       = Size(5 * 1024 * 1024)         // 5MiB
 	defaultCompressionLevel = 3
-	minChunkSize            = 1024 * 1024 // 1MB
+	defaultPruneGracePeriod = 24 * time.Hour
+	minChunkSize            = 1000 * 1000 // 1MB
 	minCompressionLevel     = 1
 	maxCompressionLevel     = 19
+	// S3 accepts a multipart upload part from 5MiB to 5GiB.
+	minS3PartSize = 5 * 1024 * 1024
+	maxS3PartSize = 5 * 1024 * 1024 * 1024
 )
 
 // Sentinel validation errors.
 var (
 	errNoConfigPath         = errors.New("config path not provided")
-	errNoAgeRecipients = errors.New(
-		"at least one age_recipient is required (generate with: age-keygen)")
+	errRecipientIsSecretKey = errors.New(
+		"an age secret key was given where a public key (age1...) belongs")
+	errRecipientNotX25519 = errors.New(
+		"not a valid recipient; only X25519 age1... public keys are supported")
 	errNoSnapshots = errors.New(
 		"at least one snapshot must be configured (see config.example.yml)")
 	errSnapshotNoPaths   = errors.New("snapshot must have at least one path")
 	errChunkSizeTooSmall = errors.New("chunk_size must be at least 1MB")
-	errBlobSizeTooSmall  = errors.New("blob_size_limit must be at least chunk_size")
+	errBlobSizeTooSmall  = errors.New(
+		"blob_size_limit must be at least the largest chunk the chunker can " +
+			"emit (chunk_size times the FastCDC size spread)")
 	errBadCompression   = errors.New("compression_level must be between 1 and 19")
+	errBadS3PartSize    = errors.New("s3.part_size must be between 5MiB and 5GiB")
 	errBadStorageScheme = errors.New(
 		"storage_url must start with s3://, file://, or rclone://")
+	errFileURLNotAbsolute = errors.New(
+		"a file:// storage_url needs an absolute path, " +
+			"as in file:///path/to/backup")
+	errNegativePruneGracePeriod = errors.New(
+		"prune_grace_period must not be negative")
 	errStorageNotConfigured = errors.New(
 		"storage not configured; set storage_url or provide s3.endpoint + " +
 			"s3.bucket + credentials")
@@ -121,6 +147,26 @@ func (c *Config) SnapshotNames() []string {
 	return names
 }
 
+// Names of the two places the age secret key can be configured, used by
+// AgeSecretKeySourceName for error messages that must not echo the value.
+//
+//nolint:gosec // G101: these are the names of the config sources, not a key
+const (
+	ageSecretKeySourceEnv    = "VAULTIK_AGE_SECRET_KEY"
+	ageSecretKeySourceConfig = "age_secret_key"
+)
+
+// AgeSecretKeySourceName returns the human name of where AgeSecretKey was
+// configured. A Config built directly (as in tests) has no recorded
+// source, so it reports the config-file field name.
+func (c *Config) AgeSecretKeySourceName() string {
+	if c.AgeSecretKeySource != "" {
+		return c.AgeSecretKeySource
+	}
+
+	return ageSecretKeySourceConfig
+}
+
 // Config represents the application configuration for Vaultik.
 // It defines all settings for backup operations, including source directories,
 // encryption recipients, storage configuration, and performance tuning parameters.
@@ -130,6 +176,11 @@ func (c *Config) SnapshotNames() []string {
 type Config struct {
 	AgeRecipients []string `yaml:"age_recipients"`
 	AgeSecretKey  string   `yaml:"age_secret_key"`
+	// AgeSecretKeySource names where AgeSecretKey was configured
+	// ("VAULTIK_AGE_SECRET_KEY" or "age_secret_key") so a later parse
+	// failure can name the source without echoing the secret value. It is
+	// set by Load and never read from or written to the config file.
+	AgeSecretKeySource string `yaml:"-"`
 	BlobSizeLimit      Size   `yaml:"blob_size_limit"`
 	ChunkSize          Size   `yaml:"chunk_size"`
 	// Exclude holds global excludes applied to all snapshots.
@@ -140,6 +191,15 @@ type Config struct {
 	Snapshots        map[string]SnapshotConfig `yaml:"snapshots"`
 	CompressionLevel int                       `yaml:"compression_level"`
 
+	// PruneGracePeriod is how old a blob no manifest lists must be before
+	// a prune deletes it, and how old every object in a directory under
+	// metadata/ with no manifest must be before a prune deletes that
+	// directory. A backup writes its manifest only once it has uploaded
+	// all its blobs and its db.zst.age, so a younger blob or directory may
+	// belong to a backup still running, on this host or another sharing
+	// the destination store.
+	PruneGracePeriod time.Duration `yaml:"prune_grace_period"`
+
 	// StorageURL specifies the storage backend using a URL format.
 	// Takes precedence over S3Config if set.
 	// Supported formats:
@@ -162,7 +222,9 @@ type S3Config struct {
 	AccessKeyID     string `yaml:"access_key_id"`
 	SecretAccessKey string `yaml:"secret_access_key"`
 	Region          string `yaml:"region"`
-	UseSSL          bool   `yaml:"use_ssl"`
+	// UseSSL selects HTTPS for a scheme-less endpoint. Omitted (nil) means
+	// the default, TLS; set it to false only to force plain HTTP.
+	UseSSL   *bool `yaml:"use_ssl"`
 	PartSize Size  `yaml:"part_size"`
 }
 
@@ -187,14 +249,48 @@ func New(path Path) (*Config, error) {
 	return cfg, nil
 }
 
-// Load reads and parses the configuration file from the specified path.
-// It applies default values for optional fields, performs environment variable
-// substitution using smartconfig, and validates the configuration.
-// The configuration file should be in YAML format. Returns an error if the file
-// cannot be read, parsed, or if validation fails.
+// Load reads the configuration file at path, parses it with Parse, warns
+// when others can read the file, and validates the configuration.
+// Returns an error if the file cannot be read, parsed, or if validation
+// fails.
 func Load(path string) (*Config, error) {
+	data, err := os.ReadFile(path) //nolint:gosec // G304: config path is operator-supplied
+	if err != nil {
+		return nil, fmt.Errorf("failed to load config file: %w", err)
+	}
+
+	cfg, err := Parse(data)
+	if err != nil {
+		return nil, err
+	}
+
+	// Check config file permissions (warn if world or group readable)
+	//nolint:gosec // G703: config path is operator-supplied by design
+	info, statErr := os.Stat(path)
+	if statErr == nil {
+		mode := info.Mode().Perm()
+		if mode&0044 != 0 { // group or world readable
+			log.Warn(cfg.readableByOthersWarning(),
+				"path", path,
+				"mode", fmt.Sprintf("%04o", mode),
+				"recommendation", "chmod 600 "+path)
+		}
+	}
+
+	err = cfg.Validate()
+	if err != nil {
+		return nil, fmt.Errorf("invalid config: %w", err)
+	}
+
+	return cfg, nil
+}
+
+// Parse reads YAML config data as Load does, without validating it. It
+// applies default values for optional fields, fills in ${...} references
+// using smartconfig, and expands ~ in paths.
+func Parse(data []byte) (*Config, error) {
 	// Load config using smartconfig for interpolation
-	sc, err := smartconfig.NewFromConfigPath(path)
+	sc, err := smartconfig.NewFromReader(bytes.NewReader(data))
 	if err != nil {
 		return nil, fmt.Errorf("failed to load config file: %w", err)
 	}
@@ -205,6 +301,8 @@ func Load(path string) (*Config, error) {
 		ChunkSize:        defaultChunkSize,
 		IndexPath:        filepath.Join(xdg.DataHome, appName, "index.sqlite"),
 		CompressionLevel: defaultCompressionLevel,
+		PruneGracePeriod: defaultPruneGracePeriod,
+		S3:               S3Config{PartSize: defaultS3PartSize},
 	}
 
 	// Convert smartconfig data to YAML then unmarshal
@@ -238,10 +336,7 @@ func Load(path string) (*Config, error) {
 		cfg.IndexPath = expandTilde(envIndexPath)
 	}
 
-	// Check for environment variable override for AgeSecretKey
-	if envAgeSecretKey := os.Getenv("VAULTIK_AGE_SECRET_KEY"); envAgeSecretKey != "" {
-		cfg.AgeSecretKey = extractAgeSecretKey(envAgeSecretKey)
-	}
+	cfg.setAgeSecretKey()
 
 	// Get hostname if not set
 	if cfg.Hostname == "" {
@@ -258,43 +353,32 @@ func Load(path string) (*Config, error) {
 		cfg.S3.Region = "us-east-1"
 	}
 
-	if cfg.S3.PartSize == 0 {
-		cfg.S3.PartSize = defaultS3PartSize
-	}
-
-	// Check config file permissions (warn if world or group readable)
-	//nolint:gosec // G703: config path is operator-supplied by design
-	info, statErr := os.Stat(path)
-	if statErr == nil {
-		mode := info.Mode().Perm()
-		if mode&0044 != 0 { // group or world readable
-			log.Warn("Config file has insecure permissions (contains S3 credentials)",
-				"path", path,
-				"mode", fmt.Sprintf("%04o", mode),
-				"recommendation", "chmod 600 "+path)
-		}
-	}
-
-	err = cfg.Validate()
-	if err != nil {
-		return nil, fmt.Errorf("invalid config: %w", err)
-	}
-
 	return cfg, nil
 }
 
 // Validate checks if the configuration is valid and complete.
 // It ensures all required fields are present and have valid values:
-// - At least one age recipient must be specified
+//   - Every age recipient must parse as an X25519 age1... public key (so a
+//     bad entry fails at load, not mid-backup); errors name the position,
+//     never the value. An empty list is accepted, because only snapshot
+//     create needs a recipient and it checks for one itself
 //   - At least one snapshot must be configured with at least one path
 //   - Storage must be configured (either storage_url or s3.* fields)
 //   - Chunk size must be at least 1MB
-// - Blob size limit must be at least the chunk size
+//   - Blob size limit must be at least the largest chunk the chunker can emit
+//     (chunk_size times chunker.ChunkSizeSpread), so a single-chunk blob never
+//     exceeds the configured limit
 //   - Compression level must be between 1 and 19
+//   - S3 part size must be between 5MiB and 5GiB, the part sizes S3 accepts
+//   - Prune grace period must not be negative
+//
 // Returns an error describing the first validation failure encountered.
 func (c *Config) Validate() error {
-	if len(c.AgeRecipients) == 0 {
-		return errNoAgeRecipients
+	for i, recipient := range c.AgeRecipients {
+		err := validateAgeRecipient(recipient)
+		if err != nil {
+			return fmt.Errorf("age_recipients[%d]: %w", i, err)
+		}
 	}
 
 	if len(c.Snapshots) == 0 {
@@ -317,8 +401,13 @@ func (c *Config) Validate() error {
 		return errChunkSizeTooSmall
 	}
 
-	if c.BlobSizeLimit.Int64() < c.ChunkSize.Int64() {
-		return errBlobSizeTooSmall
+	// The chunker can emit chunks up to chunk_size * ChunkSizeSpread, and the
+	// packer places a single such chunk into an otherwise empty blob. A limit
+	// below that bound would let a blob exceed it, so reject it.
+	largestChunk := c.ChunkSize.Int64() * chunker.ChunkSizeSpread
+	if c.BlobSizeLimit.Int64() < largestChunk {
+		return fmt.Errorf("%w: need at least %d bytes",
+			errBlobSizeTooSmall, largestChunk)
 	}
 
 	if c.CompressionLevel < minCompressionLevel ||
@@ -326,9 +415,62 @@ func (c *Config) Validate() error {
 		return errBadCompression
 	}
 
+	if c.S3.PartSize.Int64() < minS3PartSize ||
+		c.S3.PartSize.Int64() > maxS3PartSize {
+		return errBadS3PartSize
+	}
+
+	if c.PruneGracePeriod < 0 {
+		return errNegativePruneGracePeriod
+	}
+
 	return nil
 }
 
+// validateAgeRecipient parses one age_recipients entry with the age library
+// and returns a value-free error on failure. A recipient string can be
+// sensitive (an operator may paste a secret key by mistake), so neither the
+// entry nor age's own error (which quotes its input) is ever included.
+func validateAgeRecipient(recipient string) error {
+	if strings.HasPrefix(strings.ToUpper(recipient), secretKeyPrefix) {
+		return errRecipientIsSecretKey
+	}
+
+	_, err := age.ParseX25519Recipient(recipient)
+	if err != nil {
+		return errRecipientNotX25519
+	}
+
+	return nil
+}
+
+// setAgeSecretKey records the age secret key and where it came from. The
+// value is stored raw and parsed only where decryption happens
+// (internal/vaultik), so backup, list and prune keep working whatever the
+// field holds. The environment variable overrides the config-file field.
+func (c *Config) setAgeSecretKey() {
+	if c.AgeSecretKey != "" {
+		c.AgeSecretKeySource = ageSecretKeySourceConfig
+	}
+
+	if env := os.Getenv("VAULTIK_AGE_SECRET_KEY"); env != "" {
+		c.AgeSecretKey = env
+		c.AgeSecretKeySource = ageSecretKeySourceEnv
+	}
+}
+
+// readableByOthersWarning is the warning Load logs when others can read
+// the config file. It says "may contain" because the S3 credentials are
+// seen only after smartconfig has replaced any ${...} reference in the
+// file with its value, so a set credential need not be in the file.
+func (c *Config) readableByOthersWarning() string {
+	if c.S3.AccessKeyID != "" || c.S3.SecretAccessKey != "" {
+		return "Config file is readable by others and may contain S3 credentials"
+	}
+
+	return "Config file is readable by others"
+}
+
 // validateStorage validates storage configuration.
 // If StorageURL is set, it takes precedence. S3 URLs require credentials.
 // File URLs don't require any S3 configuration.
@@ -360,11 +502,18 @@ func (c *Config) validateStorage() error {
 
 // validateStorageURL validates URL-based storage configuration. File and
 // rclone URLs need no credentials; S3 URLs require the legacy s3.*
-// credential fields.
+// credential fields. A file URL must name an absolute path: a relative one
+// would resolve against the working directory, so the destination would
+// move with it.
 func (c *Config) validateStorageURL() error {
 	switch {
 	case strings.HasPrefix(c.StorageURL, "file://"):
-		// File storage doesn't need S3 credentials
+		// storage.ParseStorageURL applies the same rule, but config
+		// cannot import storage, which imports config.
+		if !strings.HasPrefix(c.StorageURL, "file:///") {
+			return errFileURLNotAbsolute
+		}
+
 		return nil
 	case strings.HasPrefix(c.StorageURL, "rclone://"):
 		// Rclone storage uses rclone's own config
@@ -385,22 +534,6 @@ func (c *Config) validateStorageURL() error {
 	}
 }
 
-// extractAgeSecretKey extracts the AGE-SECRET-KEY from the input using
-// the age library's parser, which handles comments and whitespace.
-func extractAgeSecretKey(input string) string {
-	identities, err := age.ParseIdentities(strings.NewReader(input))
-	if err != nil || len(identities) == 0 {
-		// Fall back to trimmed input if parsing fails
-		return strings.TrimSpace(input)
-	}
-	// Return the string representation of the first identity
-	if id, ok := identities[0].(*age.X25519Identity); ok {
-		return id.String()
-	}
-
-	return strings.TrimSpace(input)
-}
-
 // Module exports the config module for fx dependency injection.
 // It provides the Config type to other modules in the application.
 //

internal/config/config_test.go

@@ -1,9 +1,15 @@
-package config //nolint:testpackage // exercises unexported extractAgeSecretKey
+package config //nolint:testpackage // exercises unexported source constants
 
 import (
+	"errors"
 	"os"
 	"path/filepath"
+	"strings"
 	"testing"
+	"time"
+
+	"sneak.berlin/go/vaultik/internal/chunker"
+	"sneak.berlin/go/vaultik/internal/log"
 )
 
 const (
@@ -83,6 +89,48 @@ func TestConfigLoad(t *testing.T) {
 	}
 }
 
+// TestExampleConfigIsScrubbedAndLoads checks that the shipped
+// config.example.yml carries only neutral placeholders (no real credentials,
+// private addresses, or internal host names) and still parses.
+func TestExampleConfigIsScrubbedAndLoads(t *testing.T) {
+	t.Parallel()
+
+	examplePath := filepath.Join("..", "..", "config.example.yml")
+
+	cfg, err := Load(examplePath)
+	if err != nil {
+		t.Fatalf("Failed to load config.example.yml: %v", err)
+	}
+
+	if cfg.StorageURL != "rclone://myremote/path/to/backups" {
+		t.Errorf("Expected neutral storage_url, got '%s'", cfg.StorageURL)
+	}
+
+	//nolint:gosec // G304: examplePath is a fixed in-repo path, not user input
+	raw, err := os.ReadFile(examplePath)
+	if err != nil {
+		t.Fatalf("Failed to read config.example.yml: %v", err)
+	}
+
+	text := string(raw)
+
+	wantSubstrings := []string{
+		"YOUR_ACCESS_KEY",
+		"YOUR_SECRET_KEY",
+		"endpoint: https://",
+	}
+	for _, want := range wantSubstrings {
+		if !strings.Contains(text, want) {
+			t.Errorf("Expected config.example.yml to contain %q", want)
+		}
+	}
+
+	// A raw "http://" scheme would mean a plaintext, likely private endpoint.
+	if strings.Contains(text, "http://") {
+		t.Error("config.example.yml should not contain an http:// endpoint")
+	}
+}
+
 // TestConfigFromEnv tests loading config path from environment variable
 func TestConfigFromEnv(t *testing.T) {
 	t.Parallel()
@@ -101,53 +149,58 @@ func TestConfigFromEnv(t *testing.T) {
 	}
 }
 
-// TestExtractAgeSecretKey tests extraction of AGE-SECRET-KEY from various inputs
-func TestExtractAgeSecretKey(t *testing.T) {
+// TestValidateBlobSizeLimit checks the blob_size_limit boundary: it must be at
+// least the largest chunk the chunker can emit (chunk_size times
+// chunker.ChunkSizeSpread), because the packer places a single such chunk into
+// an otherwise empty blob. A limit between chunk_size and that bound is rejected.
+func TestValidateBlobSizeLimit(t *testing.T) {
 	t.Parallel()
 
+	const chunkSize = Size(10 * 1024 * 1024) // 10MB
+
+	largestChunk := chunkSize.Int64() * chunker.ChunkSizeSpread
+
+	newConfig := func(blobLimit Size) *Config {
+		return &Config{
+			AgeRecipients:    []string{testSneakAgePublicKey},
+			Snapshots:        map[string]SnapshotConfig{"test": {Paths: []string{"/tmp/src"}}},
+			StorageURL:       "file:///tmp/vaultik-test-store",
+			ChunkSize:        chunkSize,
+			BlobSizeLimit:    blobLimit,
+			CompressionLevel: 3,
+			S3:               S3Config{PartSize: defaultS3PartSize},
+		}
+	}
+
 	tests := []struct {
 		name      string
-		input    string
-		expected string
+		blobLimit Size
+		wantErr   bool
 	}{
 		{
-			name:     "plain key",
-			input:    testIntegrationAgePrivateKey,
-			expected: testIntegrationAgePrivateKey,
+			name:      "at chunk_size but below largest chunk is rejected",
+			blobLimit: chunkSize,
+			wantErr:   true,
 		},
 		{
-			name:     "key with trailing newline",
-			input:    testIntegrationAgePrivateKey + "\n",
-			expected: testIntegrationAgePrivateKey,
+			name:      "between chunk_size and largest chunk is rejected",
+			blobLimit: Size(chunkSize.Int64() * 2),
+			wantErr:   true,
 		},
 		{
-			name: "full age-keygen output",
-			input: "# created: 2025-01-14T12:00:00Z\n" +
-				"# public key: " + testIntegrationAgePublicKey + "\n" +
-				testIntegrationAgePrivateKey + "\n",
-			expected: testIntegrationAgePrivateKey,
+			name:      "one byte below largest chunk is rejected",
+			blobLimit: Size(largestChunk - 1),
+			wantErr:   true,
 		},
 		{
-			name: "age-keygen output with extra blank lines",
-			input: "# created: 2025-01-14T12:00:00Z\n" +
-				"# public key: " + testIntegrationAgePublicKey + "\n\n" +
-				testIntegrationAgePrivateKey + "\n\n",
-			expected: testIntegrationAgePrivateKey,
+			name:      "exactly at largest chunk is accepted",
+			blobLimit: Size(largestChunk),
+			wantErr:   false,
 		},
 		{
-			name:     "key with leading whitespace",
-			input:    "  " + testIntegrationAgePrivateKey + "  ",
-			expected: testIntegrationAgePrivateKey,
-		},
-		{
-			name:     "empty input",
-			input:    "",
-			expected: "",
-		},
-		{
-			name:     "only comments",
-			input:    "# this is a comment\n# another comment",
-			expected: "# this is a comment\n# another comment",
+			name:      "above largest chunk is accepted",
+			blobLimit: Size(largestChunk * 100),
+			wantErr:   false,
 		},
 	}
 
@@ -155,10 +208,549 @@ func TestExtractAgeSecretKey(t *testing.T) {
 		t.Run(tt.name, func(t *testing.T) {
 			t.Parallel()
 
-			result := extractAgeSecretKey(tt.input)
-			if result != tt.expected {
-				t.Errorf("extractAgeSecretKey(%q) = %q, want %q",
-					tt.input, result, tt.expected)
+			err := newConfig(tt.blobLimit).Validate()
+			if tt.wantErr {
+				if !errors.Is(err, errBlobSizeTooSmall) {
+					t.Fatalf("Validate() error = %v, want errBlobSizeTooSmall", err)
+				}
+
+				return
+			}
+
+			if err != nil {
+				t.Fatalf("Validate() unexpected error: %v", err)
+			}
+		})
+	}
+}
+
+// TestValidateS3PartSize checks that s3.part_size is held to the part sizes
+// S3 accepts, 5MiB to 5GiB, by changing only the part size of the test
+// config. "5MB" in the config file is 5,000,000 bytes, below the minimum.
+func TestValidateS3PartSize(t *testing.T) {
+	t.Parallel()
+
+	base, err := Load(os.Getenv("VAULTIK_CONFIG"))
+	if err != nil {
+		t.Fatalf("Failed to load config: %v", err)
+	}
+
+	tests := []struct {
+		name     string
+		partSize Size
+		wantErr  bool
+	}{
+		{
+			name:     "5MB is rejected",
+			partSize: 5_000_000,
+			wantErr:  true,
+		},
+		{
+			name:     "one byte below 5MiB is rejected",
+			partSize: minS3PartSize - 1,
+			wantErr:  true,
+		},
+		{
+			name:     "5MiB is accepted",
+			partSize: minS3PartSize,
+			wantErr:  false,
+		},
+		{
+			name:     "5GiB is accepted",
+			partSize: maxS3PartSize,
+			wantErr:  false,
+		},
+		{
+			name:     "one byte above 5GiB is rejected",
+			partSize: maxS3PartSize + 1,
+			wantErr:  true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			cfg := *base
+			cfg.S3.PartSize = tt.partSize
+
+			err := cfg.Validate()
+			if tt.wantErr {
+				if !errors.Is(err, errBadS3PartSize) {
+					t.Fatalf("Validate() error = %v, want errBadS3PartSize", err)
+				}
+
+				return
+			}
+
+			if err != nil {
+				t.Fatalf("Validate() unexpected error: %v", err)
+			}
+		})
+	}
+}
+
+// TestLoadS3PartSize checks that a config file without s3.part_size loads
+// with the 5MiB default, and that an explicit 0 fails at load like any other
+// part size S3 refuses.
+func TestLoadS3PartSize(t *testing.T) {
+	t.Parallel()
+
+	const withoutPartSize = "snapshots:\n" +
+		"  test:\n" +
+		"    paths: [/tmp/vaultik-test-source]\n" +
+		"storage_url: file:///tmp/vaultik-test-storage\n"
+
+	writeConfig := func(t *testing.T, text string) string {
+		t.Helper()
+
+		path := filepath.Join(t.TempDir(), "config.yml")
+
+		err := os.WriteFile(path, []byte(text), 0o600)
+		if err != nil {
+			t.Fatalf("write config: %v", err)
+		}
+
+		return path
+	}
+
+	t.Run("absent loads as 5MiB", func(t *testing.T) {
+		t.Parallel()
+
+		cfg, err := Load(writeConfig(t, withoutPartSize))
+		if err != nil {
+			t.Fatalf("Load() unexpected error: %v", err)
+		}
+
+		if cfg.S3.PartSize != defaultS3PartSize {
+			t.Errorf("s3.part_size = %d, want %d",
+				cfg.S3.PartSize, defaultS3PartSize)
+		}
+	})
+
+	t.Run("0 is rejected", func(t *testing.T) {
+		t.Parallel()
+
+		_, err := Load(writeConfig(t, withoutPartSize+"s3:\n  part_size: 0\n"))
+		if !errors.Is(err, errBadS3PartSize) {
+			t.Fatalf("Load() error = %v, want errBadS3PartSize", err)
+		}
+	})
+}
+
+// TestDocumentedSizeDefaults checks that the defaults are the documented
+// chunk_size 10MB and blob_size_limit 10GB as the parser reads those values
+// (1000-based), so writing a default out in the config file changes nothing.
+func TestDocumentedSizeDefaults(t *testing.T) {
+	t.Parallel()
+
+	cfg, err := Parse([]byte("chunk_size: 10MB\nblob_size_limit: 10GB\n"))
+	if err != nil {
+		t.Fatalf("Parse() unexpected error: %v", err)
+	}
+
+	if cfg.ChunkSize != defaultChunkSize {
+		t.Errorf("chunk_size: 10MB is %d bytes, the default is %d",
+			cfg.ChunkSize, defaultChunkSize)
+	}
+
+	if cfg.BlobSizeLimit != defaultBlobSizeLimit {
+		t.Errorf("blob_size_limit: 10GB is %d bytes, the default is %d",
+			cfg.BlobSizeLimit, defaultBlobSizeLimit)
+	}
+}
+
+// TestDocumentedSizeExamples checks the 1MB minimum chunk_size and that
+// every size the README, config.example.yml and the config init template
+// list as an example is accepted.
+func TestDocumentedSizeExamples(t *testing.T) {
+	t.Parallel()
+
+	const withoutSizes = "snapshots:\n" +
+		"  test:\n" +
+		"    paths: [/tmp/vaultik-test-source]\n" +
+		"storage_url: file:///tmp/vaultik-test-storage\n"
+
+	tests := []struct {
+		setting string
+		wantErr error
+	}{
+		{"chunk_size: 1MB", nil},
+		{"chunk_size: 999999", errChunkSizeTooSmall},
+		{"chunk_size: 5M", nil},
+		{"chunk_size: 10M", nil},
+		{"chunk_size: 10MB", nil},
+		{"chunk_size: 64MiB", nil},
+		{"chunk_size: 1GB", nil},
+		// Four times the default chunk_size, the documented minimum.
+		{"blob_size_limit: 40MB", nil},
+		{"blob_size_limit: 500MB", nil},
+		{"blob_size_limit: 1GB", nil},
+		{"blob_size_limit: 1GiB", nil},
+		{"blob_size_limit: 10G", nil},
+		{"blob_size_limit: 10GB", nil},
+		{"s3: {part_size: 5MiB}", nil},
+		{"s3: {part_size: 10MB}", nil},
+		{"s3: {part_size: 16MiB}", nil},
+		{"s3: {part_size: 100MiB}", nil},
+		{"s3: {part_size: 5GiB}", nil},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.setting, func(t *testing.T) {
+			t.Parallel()
+
+			cfg, err := Parse([]byte(withoutSizes + tt.setting + "\n"))
+			if err != nil {
+				t.Fatalf("Parse() unexpected error: %v", err)
+			}
+
+			err = cfg.Validate()
+			if !errors.Is(err, tt.wantErr) {
+				t.Fatalf("Validate() error = %v, want %v", err, tt.wantErr)
+			}
+		})
+	}
+}
+
+// TestValidatePruneGracePeriod checks that a negative prune_grace_period
+// is refused rather than acting like 0s, which deletes unreferenced blobs
+// and directories under metadata/ with no manifest at once, and that 0s
+// and the documented examples are accepted.
+func TestValidatePruneGracePeriod(t *testing.T) {
+	t.Parallel()
+
+	const withoutGracePeriod = "snapshots:\n" +
+		"  test:\n" +
+		"    paths: [/tmp/vaultik-test-source]\n" +
+		"storage_url: file:///tmp/vaultik-test-storage\n"
+
+	tests := []struct {
+		setting string
+		wantErr error
+	}{
+		{"prune_grace_period: 0s", nil},
+		{"prune_grace_period: 90m", nil},
+		{"prune_grace_period: 36h", nil},
+		{"prune_grace_period: -1h", errNegativePruneGracePeriod},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.setting, func(t *testing.T) {
+			t.Parallel()
+
+			cfg, err := Parse([]byte(withoutGracePeriod + tt.setting + "\n"))
+			if err != nil {
+				t.Fatalf("Parse() unexpected error: %v", err)
+			}
+
+			err = cfg.Validate()
+			if !errors.Is(err, tt.wantErr) {
+				t.Fatalf("Validate() error = %v, want %v", err, tt.wantErr)
+			}
+		})
+	}
+}
+
+// TestLoadPruneGracePeriod checks that a config file without
+// prune_grace_period loads with the documented 24h. The value is written out
+// here rather than taken from defaultPruneGracePeriod, so that a change of the
+// default fails this test.
+func TestLoadPruneGracePeriod(t *testing.T) {
+	t.Parallel()
+
+	const withoutGracePeriod = "snapshots:\n" +
+		"  test:\n" +
+		"    paths: [/tmp/vaultik-test-source]\n" +
+		"storage_url: file:///tmp/vaultik-test-storage\n"
+
+	path := filepath.Join(t.TempDir(), "config.yml")
+
+	err := os.WriteFile(path, []byte(withoutGracePeriod), 0o600)
+	if err != nil {
+		t.Fatalf("write config: %v", err)
+	}
+
+	cfg, err := Load(path)
+	if err != nil {
+		t.Fatalf("Load() unexpected error: %v", err)
+	}
+
+	if cfg.PruneGracePeriod != 24*time.Hour {
+		t.Errorf("prune_grace_period = %s, want 24h", cfg.PruneGracePeriod)
+	}
+}
+
+// TestLoadRejectsRelativeFileURL checks that a file:// storage_url with a
+// relative path fails at load. Accepted, it would resolve against the
+// working directory, so a backup run from another directory would write
+// to a different destination and still succeed.
+func TestLoadRejectsRelativeFileURL(t *testing.T) {
+	t.Parallel()
+
+	const relativeConfig = "snapshots:\n" +
+		"  test:\n" +
+		"    paths: [/tmp/vaultik-test-source]\n" +
+		"storage_url: file://dest\n"
+
+	path := filepath.Join(t.TempDir(), "config.yml")
+
+	err := os.WriteFile(path, []byte(relativeConfig), 0o600)
+	if err != nil {
+		t.Fatalf("write config: %v", err)
+	}
+
+	_, err = Load(path)
+	if !errors.Is(err, errFileURLNotAbsolute) {
+		t.Fatalf("Load() error = %v, want errFileURLNotAbsolute", err)
+	}
+
+	if !strings.Contains(err.Error(), "absolute path") {
+		t.Errorf("error %q does not say to use an absolute path", err.Error())
+	}
+}
+
+// TestValidateAgeRecipients checks that recipients are parsed at config load
+// (a bad entry fails immediately, not mid-backup) and that no invalid entry —
+// least of all a pasted secret key — is echoed in the error. An empty list
+// loads, because only snapshot create needs a recipient.
+func TestValidateAgeRecipients(t *testing.T) {
+	t.Parallel()
+
+	baseConfig := func(recipients []string) *Config {
+		return &Config{
+			AgeRecipients:    recipients,
+			Snapshots:        map[string]SnapshotConfig{"test": {Paths: []string{"/tmp/src"}}},
+			StorageURL:       "file:///tmp/vaultik-test-store",
+			ChunkSize:        Size(10 * 1024 * 1024),
+			BlobSizeLimit:    Size(10 * 1024 * 1024 * 1024),
+			CompressionLevel: 3,
+			S3:               S3Config{PartSize: defaultS3PartSize},
+		}
+	}
+
+	tests := []struct {
+		name       string
+		recipients []string
+		wantErr    bool
+	}{
+		{
+			name:       "no recipients is accepted",
+			recipients: nil,
+			wantErr:    false,
+		},
+		{
+			name:       "placeholder recipient is rejected",
+			recipients: []string{"age1REPLACE_WITH_YOUR_PUBLIC_KEY"},
+			wantErr:    true,
+		},
+		{
+			name:       "ssh-ed25519 recipient is rejected",
+			recipients: []string{"ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIexamplekeydata"},
+			wantErr:    true,
+		},
+		{
+			name:       "truncated age1 string is rejected",
+			recipients: []string{"age1short"},
+			wantErr:    true,
+		},
+		{
+			name:       "secret key passed as recipient is rejected",
+			recipients: []string{testIntegrationAgePrivateKey},
+			wantErr:    true,
+		},
+		{
+			name:       "two valid recipients are accepted",
+			recipients: []string{testSneakAgePublicKey, testIntegrationAgePublicKey},
+			wantErr:    false,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			err := baseConfig(tt.recipients).Validate()
+			if !tt.wantErr {
+				if err != nil {
+					t.Fatalf("Validate() unexpected error: %v", err)
+				}
+
+				return
+			}
+
+			if err == nil {
+				t.Fatal("Validate() returned nil, want error")
+			}
+
+			// The entry itself must never appear in the error, since a
+			// recipient string can be a secret key.
+			for _, recipient := range tt.recipients {
+				if strings.Contains(err.Error(), recipient) {
+					t.Fatalf("Validate() error echoed the recipient value: %v", err)
+				}
+			}
+		})
+	}
+}
+
+// TestAgeSecretKeySourceName checks the name reported for the configured
+// age secret key: the recorded source when Load set one, and the
+// config-file field name for a Config built directly (as in tests).
+func TestAgeSecretKeySourceName(t *testing.T) {
+	t.Parallel()
+
+	tests := []struct {
+		name   string
+		source string
+		want   string
+	}{
+		{
+			name:   "unset defaults to config field",
+			source: "",
+			want:   ageSecretKeySourceConfig,
+		},
+		{
+			name:   "environment source",
+			source: ageSecretKeySourceEnv,
+			want:   ageSecretKeySourceEnv,
+		},
+		{
+			name:   "config-file source",
+			source: ageSecretKeySourceConfig,
+			want:   ageSecretKeySourceConfig,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			t.Parallel()
+
+			cfg := &Config{AgeSecretKeySource: tt.source}
+			if got := cfg.AgeSecretKeySourceName(); got != tt.want {
+				t.Errorf("AgeSecretKeySourceName() = %q, want %q", got, tt.want)
+			}
+		})
+	}
+}
+
+// loadReadableConfig writes configYAML to a file that others can read,
+// loads it, and returns what the logger wrote to stderr meanwhile. The
+// logger writes to the os.Stderr it finds when it is initialized, so
+// os.Stderr is pointed at a file first. Not parallel-safe: os.Stderr and
+// the logger are process-global.
+func loadReadableConfig(t *testing.T, configYAML string) string {
+	t.Helper()
+
+	dir := t.TempDir()
+	configPath := filepath.Join(dir, "config.yml")
+	stderrPath := filepath.Join(dir, "stderr")
+
+	err := os.WriteFile(configPath, []byte(configYAML), 0o600)
+	if err != nil {
+		t.Fatalf("writing config: %v", err)
+	}
+
+	//nolint:gosec // G302: the test needs a config file others can read
+	err = os.Chmod(configPath, 0o644)
+	if err != nil {
+		t.Fatalf("chmod config: %v", err)
+	}
+
+	stderrFile, err := os.Create(stderrPath) //nolint:gosec // G304: test temp path
+	if err != nil {
+		t.Fatalf("creating stderr file: %v", err)
+	}
+
+	previous := os.Stderr
+	os.Stderr = stderrFile
+
+	log.Initialize(log.Config{})
+
+	_, loadErr := Load(configPath)
+
+	os.Stderr = previous
+
+	log.Initialize(log.Config{})
+
+	_ = stderrFile.Close()
+
+	if loadErr != nil {
+		t.Fatalf("Load() error = %v", loadErr)
+	}
+
+	captured, err := os.ReadFile(stderrPath) //nolint:gosec // G304: test temp path
+	if err != nil {
+		t.Fatalf("reading stderr file: %v", err)
+	}
+
+	return string(captured)
+}
+
+// TestLoadWarnsReadableConfigWithoutS3Credentials checks that a config
+// file others can read, holding no S3 credentials, is warned about
+// without a claim that it holds them.
+//
+//nolint:paralleltest // loadReadableConfig replaces os.Stderr
+func TestLoadWarnsReadableConfigWithoutS3Credentials(t *testing.T) {
+	stderr := loadReadableConfig(t, `
+storage_url: file:///var/backups/vaultik
+snapshots:
+  home:
+    paths:
+      - /home
+`)
+
+	if !strings.Contains(stderr, "Config file is readable by others") {
+		t.Errorf("expected a warning that the file is readable by others, got %q",
+			stderr)
+	}
+
+	if strings.Contains(stderr, "S3 credentials") {
+		t.Errorf("warning names S3 credentials the file does not set: %q", stderr)
+	}
+}
+
+// TestLoadWarnsReadableConfigWithS3Credentials checks that a config file
+// others can read and that sets S3 credentials, as values or as ${ENV:...}
+// references, is warned about as one that may contain them.
+//
+//nolint:paralleltest // loadReadableConfig replaces os.Stderr
+func TestLoadWarnsReadableConfigWithS3Credentials(t *testing.T) {
+	t.Setenv("VAULTIK_TEST_ACCESS_KEY_ID", "test-access-key")
+	t.Setenv("VAULTIK_TEST_SECRET_ACCESS_KEY", "test-secret-key")
+
+	configs := map[string]string{
+		"values": `
+storage_url: s3://bucket/prefix?endpoint=s3.example.com
+s3:
+  access_key_id: test-access-key
+  secret_access_key: test-secret-key
+snapshots:
+  home:
+    paths:
+      - /home
+`,
+		"references": `
+storage_url: s3://bucket/prefix?endpoint=s3.example.com
+s3:
+  access_key_id: ${ENV:VAULTIK_TEST_ACCESS_KEY_ID}
+  secret_access_key: ${ENV:VAULTIK_TEST_SECRET_ACCESS_KEY}
+snapshots:
+  home:
+    paths:
+      - /home
+`,
+	}
+
+	for name, configYAML := range configs {
+		t.Run(name, func(t *testing.T) {
+			stderr := loadReadableConfig(t, configYAML)
+
+			if !strings.Contains(stderr,
+				"Config file is readable by others and may contain S3 credentials") {
+				t.Errorf("expected a warning naming the S3 credentials, got %q",
+					stderr)
 			}
 		})
 	}

internal/config/size.go

@@ -16,8 +16,6 @@ var (
 // Size represents a byte size that can be specified in configuration files.
 // It can unmarshal from both numeric values (interpreted as bytes) and
 // human-readable strings like "10MB", "2.5GB", or "1TB".
-//
-//nolint:recvcheck // UnmarshalYAML requires a pointer; String/Int64 are value reads
 type Size int64
 
 // UnmarshalYAML implements yaml.Unmarshaler for Size, allowing it to be

internal/crypto/encryption.go

@@ -1,224 +0,0 @@
-// Package crypto provides thread-safe age encryption and decryption
-// helpers used to protect blob and metadata content.
-package crypto //nolint:revive,nolintlint // stdlib crypto unused; see #76
-
-import (
-	"bytes"
-	"errors"
-	"fmt"
-	"io"
-	"sync"
-
-	"filippo.io/age"
-	"go.uber.org/fx"
-)
-
-// ErrNoRecipients is returned when an encryptor is created or updated
-// without any recipient public keys.
-var ErrNoRecipients = errors.New("at least one recipient is required")
-
-// Encryptor provides thread-safe encryption using the age encryption library.
-// It supports encrypting data for multiple recipients simultaneously, allowing
-// any of the corresponding private keys to decrypt the data. This is useful
-// for backup scenarios where multiple parties should be able to decrypt the data.
-type Encryptor struct {
-	recipients []age.Recipient
-	mu         sync.RWMutex
-}
-
-// NewEncryptor creates a new encryptor with the given age public keys.
-// Each public key should be a valid age X25519 recipient string (e.g., "age1...")
-// At least one recipient must be provided. Returns an error if any of the
-// public keys are invalid or if no recipients are specified.
-func NewEncryptor(publicKeys []string) (*Encryptor, error) {
-	if len(publicKeys) == 0 {
-		return nil, ErrNoRecipients
-	}
-
-	recipients := make([]age.Recipient, 0, len(publicKeys))
-	for _, key := range publicKeys {
-		recipient, err := age.ParseX25519Recipient(key)
-		if err != nil {
-			return nil, fmt.Errorf("parsing age recipient %s: %w", key, err)
-		}
-
-		recipients = append(recipients, recipient)
-	}
-
-	return &Encryptor{
-		recipients: recipients,
-	}, nil
-}
-
-// Encrypt encrypts data using age encryption for all configured recipients.
-// The encrypted data can be decrypted by any of the corresponding private keys.
-// This method is suitable for small to medium amounts of data that fit in memory.
-// For large data streams, use EncryptStream or EncryptWriter instead.
-func (e *Encryptor) Encrypt(data []byte) ([]byte, error) {
-	e.mu.RLock()
-	recipients := e.recipients
-	e.mu.RUnlock()
-
-	var buf bytes.Buffer
-
-	// Create encrypted writer for all recipients
-	w, err := age.Encrypt(&buf, recipients...)
-	if err != nil {
-		return nil, fmt.Errorf("creating encrypted writer: %w", err)
-	}
-
-	// Write data
-	_, err = w.Write(data)
-	if err != nil {
-		return nil, fmt.Errorf("writing encrypted data: %w", err)
-	}
-
-	// Close to flush
-	err = w.Close()
-	if err != nil {
-		return nil, fmt.Errorf("closing encrypted writer: %w", err)
-	}
-
-	return buf.Bytes(), nil
-}
-
-// EncryptStream encrypts data from reader to writer using age encryption.
-// This method is suitable for encrypting large files or streams as it processes
-// data in a streaming fashion without loading everything into memory.
-// The encrypted data is written directly to the destination writer.
-func (e *Encryptor) EncryptStream(dst io.Writer, src io.Reader) error {
-	e.mu.RLock()
-	recipients := e.recipients
-	e.mu.RUnlock()
-
-	// Create encrypted writer for all recipients
-	w, err := age.Encrypt(dst, recipients...)
-	if err != nil {
-		return fmt.Errorf("creating encrypted writer: %w", err)
-	}
-
-	// Copy data
-	_, err = io.Copy(w, src)
-	if err != nil {
-		return fmt.Errorf("copying encrypted data: %w", err)
-	}
-
-	// Close to flush
-	err = w.Close()
-	if err != nil {
-		return fmt.Errorf("closing encrypted writer: %w", err)
-	}
-
-	return nil
-}
-
-// EncryptWriter creates a writer that encrypts data written to it.
-// All data written to the returned WriteCloser will be encrypted and written
-// to the destination writer. The caller must call Close() on the returned
-// writer to ensure all encrypted data is properly flushed and finalized.
-// This is useful for integrating encryption into existing writer-based pipelines.
-func (e *Encryptor) EncryptWriter(dst io.Writer) (io.WriteCloser, error) {
-	e.mu.RLock()
-	recipients := e.recipients
-	e.mu.RUnlock()
-
-	// Create encrypted writer for all recipients
-	w, err := age.Encrypt(dst, recipients...)
-	if err != nil {
-		return nil, fmt.Errorf("creating encrypted writer: %w", err)
-	}
-
-	return w, nil
-}
-
-// UpdateRecipients updates the recipients for future encryption operations.
-// This method is thread-safe and can be called while other encryption operations
-// are in progress. Existing encryption operations will continue with the old
-// recipients. At least one recipient must be provided. Returns an error if any
-// of the public keys are invalid or if no recipients are specified.
-func (e *Encryptor) UpdateRecipients(publicKeys []string) error {
-	if len(publicKeys) == 0 {
-		return ErrNoRecipients
-	}
-
-	recipients := make([]age.Recipient, 0, len(publicKeys))
-	for _, key := range publicKeys {
-		recipient, err := age.ParseX25519Recipient(key)
-		if err != nil {
-			return fmt.Errorf("parsing age recipient %s: %w", key, err)
-		}
-
-		recipients = append(recipients, recipient)
-	}
-
-	e.mu.Lock()
-	e.recipients = recipients
-	e.mu.Unlock()
-
-	return nil
-}
-
-// Decryptor provides thread-safe decryption using the age encryption library.
-// It uses a private key to decrypt data that was encrypted for the corresponding
-// public key.
-type Decryptor struct {
-	identity age.Identity
-	mu       sync.RWMutex
-}
-
-// NewDecryptor creates a new decryptor with the given age private key.
-// The private key should be a valid age X25519 identity string.
-// Returns an error if the private key is invalid.
-func NewDecryptor(privateKey string) (*Decryptor, error) {
-	identity, err := age.ParseX25519Identity(privateKey)
-	if err != nil {
-		return nil, fmt.Errorf("parsing age identity: %w", err)
-	}
-
-	return &Decryptor{
-		identity: identity,
-	}, nil
-}
-
-// Decrypt decrypts data using age decryption.
-// This method is suitable for small to medium amounts of data that fit in memory.
-// For large data streams, use DecryptStream instead.
-func (d *Decryptor) Decrypt(data []byte) ([]byte, error) {
-	d.mu.RLock()
-	identity := d.identity
-	d.mu.RUnlock()
-
-	r, err := age.Decrypt(bytes.NewReader(data), identity)
-	if err != nil {
-		return nil, fmt.Errorf("creating decrypted reader: %w", err)
-	}
-
-	decrypted, err := io.ReadAll(r)
-	if err != nil {
-		return nil, fmt.Errorf("reading decrypted data: %w", err)
-	}
-
-	return decrypted, nil
-}
-
-// DecryptStream returns a reader that decrypts data from the provided reader.
-// This method is suitable for decrypting large files or streams as it processes
-// data in a streaming fashion without loading everything into memory.
-// The caller should close the input reader when done.
-func (d *Decryptor) DecryptStream(src io.Reader) (io.Reader, error) {
-	d.mu.RLock()
-	identity := d.identity
-	d.mu.RUnlock()
-
-	r, err := age.Decrypt(src, identity)
-	if err != nil {
-		return nil, fmt.Errorf("creating decrypted reader: %w", err)
-	}
-
-	return r, nil
-}
-
-// Module exports the crypto module for fx dependency injection.
-//
-//nolint:gochecknoglobals // fx module definitions are package globals
-var Module = fx.Module("crypto")

internal/crypto/encryption_test.go

@@ -1,178 +0,0 @@
-package crypto_test
-
-import (
-	"bytes"
-	"testing"
-
-	"filippo.io/age"
-	"sneak.berlin/go/vaultik/internal/crypto"
-)
-
-func TestEncryptor(t *testing.T) {
-	t.Parallel()
-
-	// Generate a test key pair
-	identity, err := age.GenerateX25519Identity()
-	if err != nil {
-		t.Fatalf("failed to generate identity: %v", err)
-	}
-
-	publicKey := identity.Recipient().String()
-
-	// Create encryptor
-	enc, err := crypto.NewEncryptor([]string{publicKey})
-	if err != nil {
-		t.Fatalf("failed to create encryptor: %v", err)
-	}
-
-	// Test data
-	plaintext := []byte("Hello, World! This is a test message.")
-
-	// Encrypt
-	ciphertext, err := enc.Encrypt(plaintext)
-	if err != nil {
-		t.Fatalf("failed to encrypt: %v", err)
-	}
-
-	// Verify it's actually encrypted (should be larger and different)
-	if bytes.Equal(plaintext, ciphertext) {
-		t.Error("ciphertext equals plaintext")
-	}
-
-	// Decrypt to verify
-	r, err := age.Decrypt(bytes.NewReader(ciphertext), identity)
-	if err != nil {
-		t.Fatalf("failed to decrypt: %v", err)
-	}
-
-	var decrypted bytes.Buffer
-
-	_, err = decrypted.ReadFrom(r)
-	if err != nil {
-		t.Fatalf("failed to read decrypted data: %v", err)
-	}
-
-	if !bytes.Equal(plaintext, decrypted.Bytes()) {
-		t.Error("decrypted data doesn't match original")
-	}
-}
-
-func TestEncryptorMultipleRecipients(t *testing.T) {
-	t.Parallel()
-
-	// Generate three test key pairs
-	identity1, err := age.GenerateX25519Identity()
-	if err != nil {
-		t.Fatalf("failed to generate identity1: %v", err)
-	}
-
-	identity2, err := age.GenerateX25519Identity()
-	if err != nil {
-		t.Fatalf("failed to generate identity2: %v", err)
-	}
-
-	identity3, err := age.GenerateX25519Identity()
-	if err != nil {
-		t.Fatalf("failed to generate identity3: %v", err)
-	}
-
-	publicKeys := []string{
-		identity1.Recipient().String(),
-		identity2.Recipient().String(),
-		identity3.Recipient().String(),
-	}
-
-	// Create encryptor with multiple recipients
-	enc, err := crypto.NewEncryptor(publicKeys)
-	if err != nil {
-		t.Fatalf("failed to create encryptor: %v", err)
-	}
-
-	// Test data
-	plaintext := []byte("Secret message for multiple recipients")
-
-	// Encrypt
-	ciphertext, err := enc.Encrypt(plaintext)
-	if err != nil {
-		t.Fatalf("failed to encrypt: %v", err)
-	}
-
-	// Verify each recipient can decrypt
-	identities := []age.Identity{identity1, identity2, identity3}
-	for i, identity := range identities {
-		r, err := age.Decrypt(bytes.NewReader(ciphertext), identity)
-		if err != nil {
-			t.Fatalf("recipient %d failed to decrypt: %v", i+1, err)
-		}
-
-		var decrypted bytes.Buffer
-
-		_, err = decrypted.ReadFrom(r)
-		if err != nil {
-			t.Fatalf("recipient %d failed to read decrypted data: %v", i+1, err)
-		}
-
-		if !bytes.Equal(plaintext, decrypted.Bytes()) {
-			t.Errorf("recipient %d: decrypted data doesn't match original", i+1)
-		}
-	}
-}
-
-func TestEncryptorUpdateRecipients(t *testing.T) {
-	t.Parallel()
-
-	// Generate two identities
-	identity1, _ := age.GenerateX25519Identity()
-	identity2, _ := age.GenerateX25519Identity()
-
-	publicKey1 := identity1.Recipient().String()
-	publicKey2 := identity2.Recipient().String()
-
-	// Create encryptor with first key
-	enc, err := crypto.NewEncryptor([]string{publicKey1})
-	if err != nil {
-		t.Fatalf("failed to create encryptor: %v", err)
-	}
-
-	// Encrypt with first key
-	plaintext := []byte("test data")
-
-	ciphertext1, err := enc.Encrypt(plaintext)
-	if err != nil {
-		t.Fatalf("failed to encrypt: %v", err)
-	}
-
-	// Update to second key
-	err = enc.UpdateRecipients([]string{publicKey2})
-	if err != nil {
-		t.Fatalf("failed to update recipients: %v", err)
-	}
-
-	// Encrypt with second key
-	ciphertext2, err := enc.Encrypt(plaintext)
-	if err != nil {
-		t.Fatalf("failed to encrypt: %v", err)
-	}
-
-	// First ciphertext should only decrypt with first identity
-	_, err = age.Decrypt(bytes.NewReader(ciphertext1), identity1)
-	if err != nil {
-		t.Error("failed to decrypt with identity1")
-	}
-
-	_, err = age.Decrypt(bytes.NewReader(ciphertext1), identity2)
-	if err == nil {
-		t.Error("should not decrypt with identity2")
-	}
-
-	// Second ciphertext should only decrypt with second identity
-	_, err = age.Decrypt(bytes.NewReader(ciphertext2), identity2)
-	if err != nil {
-		t.Error("failed to decrypt with identity2")
-	}
-
-	_, err = age.Decrypt(bytes.NewReader(ciphertext2), identity1)
-	if err == nil {
-		t.Error("should not decrypt with identity1")
-	}
-}

internal/database/blob_chunks.go

@@ -58,12 +58,7 @@ func (r *BlobChunkRepository) GetByBlobID(
 		return nil, fmt.Errorf("querying blob chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var blobChunks []*BlobChunk
 

internal/database/blobs.go

@@ -83,12 +83,7 @@ func (r *BlobRepository) GetAll(ctx context.Context) (map[string]*Blob, error) {
 		return nil, fmt.Errorf("querying blobs: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	out := make(map[string]*Blob)
 
@@ -159,13 +154,16 @@ func (r *BlobRepository) UpdateFinished(
 	return nil
 }
 
-// UpdateUploaded marks a blob as uploaded
+// UpdateUploaded marks a blob as uploaded and records the size of the
+// stored object as its compressed size. That size can differ from the
+// local copy's when an identical blob was already stored with another
+// compression level or other recipients.
 func (r *BlobRepository) UpdateUploaded(
-	ctx context.Context, tx *sql.Tx, id string,
+	ctx context.Context, tx *sql.Tx, id string, storedSize int64,
 ) error {
 	query := `
 		UPDATE blobs
-		SET uploaded_ts = ?
+		SET uploaded_ts = ?, compressed_size = ?
 		WHERE id = ?
 	`
 
@@ -173,9 +171,9 @@ func (r *BlobRepository) UpdateUploaded(
 
 	var err error
 	if tx != nil {
-		_, err = tx.ExecContext(ctx, query, now, id)
+		_, err = tx.ExecContext(ctx, query, now, storedSize, id)
 	} else {
-		_, err = r.db.ExecWithLog(ctx, query, now, id)
+		_, err = r.db.ExecWithLog(ctx, query, now, storedSize, id)
 	}
 
 	if err != nil {
@@ -208,6 +206,30 @@ func (r *BlobRepository) DeleteOrphaned(ctx context.Context) error {
 	return nil
 }
 
+// DeleteUnuploaded deletes blob rows whose upload never completed
+// (uploaded_ts IS NULL) and returns how many were removed. Their
+// blob_chunks rows are removed by the ON DELETE CASCADE foreign key.
+// A blob is only ever attached to a snapshot once its upload has been
+// recorded, so an un-uploaded blob is never referenced by a completed
+// snapshot: dropping it discards chunk rows that point at data which
+// was never stored remotely, so the affected content is re-chunked and
+// re-uploaded on the next run.
+func (r *BlobRepository) DeleteUnuploaded(ctx context.Context) (int64, error) {
+	query := `DELETE FROM blobs WHERE uploaded_ts IS NULL`
+
+	result, err := r.db.ExecWithLog(ctx, query)
+	if err != nil {
+		return 0, fmt.Errorf("deleting un-uploaded blobs: %w", err)
+	}
+
+	rowsAffected, _ := result.RowsAffected()
+	if rowsAffected > 0 {
+		log.Debug("Deleted un-uploaded blobs", "count", rowsAffected)
+	}
+
+	return rowsAffected, nil
+}
+
 // getOne fetches a single blob row matched on the given column, or
 // (nil, nil) when no row matches.
 func (r *BlobRepository) getOne(

internal/database/blobs_test.go

@@ -122,8 +122,8 @@ func TestBlobRepositoryUpdates(t *testing.T) {
 		t.Errorf("expected compressed size 500, got %d", updated.CompressedSize)
 	}
 
-	// Test UpdateUploaded
-	err = repo.UpdateUploaded(ctx, nil, blob.ID.String())
+	// Test UpdateUploaded, with a stored object larger than the local copy
+	err = repo.UpdateUploaded(ctx, nil, blob.ID.String(), 600)
 	if err != nil {
 		t.Fatalf("failed to update blob as uploaded: %v", err)
 	}
@@ -141,6 +141,10 @@ func TestBlobRepositoryUpdates(t *testing.T) {
 	if uploaded.UploadedTS.Before(now.Add(-1 * time.Second)) {
 		t.Error("uploaded timestamp should be around test time")
 	}
+
+	if uploaded.CompressedSize != 600 {
+		t.Errorf("expected the stored size 600, got %d", uploaded.CompressedSize)
+	}
 }
 
 func TestBlobRepositoryDuplicate(t *testing.T) {

internal/database/chunk_files.go

@@ -61,12 +61,7 @@ func (r *ChunkFileRepository) GetByChunkHash(
 		return nil, fmt.Errorf("querying chunk files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanChunkFiles(rows)
 }
@@ -87,12 +82,7 @@ func (r *ChunkFileRepository) GetByFilePath(
 		return nil, fmt.Errorf("querying chunk files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanChunkFiles(rows)
 }
@@ -112,12 +102,7 @@ func (r *ChunkFileRepository) GetByFileID(
 		return nil, fmt.Errorf("querying chunk files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanChunkFiles(rows)
 }
@@ -220,7 +205,7 @@ func (r *ChunkFileRepository) CreateBatch(
 				cf.ChunkHash.String(), cf.FileID.String(), cf.FileOffset, cf.Length)
 		}
 
-		query += querySb183.String() //nolint:gosec // G202: appends "?" placeholders only
+		query += querySb183.String()
 
 		query += " ON CONFLICT(chunk_hash, file_id) DO NOTHING"
 

internal/database/chunks.go

@@ -98,7 +98,7 @@ func (r *ChunkRepository) GetByHashes(
 		args[i] = hash
 	}
 
-	query += querySb75.String() //nolint:gosec // G202: appends "?" placeholders only
+	query += querySb75.String()
 
 	query += ") ORDER BY chunk_hash"
 
@@ -107,12 +107,7 @@ func (r *ChunkRepository) GetByHashes(
 		return nil, fmt.Errorf("querying chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var chunks []*Chunk
 
@@ -152,12 +147,7 @@ func (r *ChunkRepository) ListUnpacked(
 		return nil, fmt.Errorf("querying unpacked chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var chunks []*Chunk
 

internal/database/chunks_ext.go

@@ -7,23 +7,38 @@ import (
 
 // List returns every chunk in the index, ordered by chunk hash.
 func (r *ChunkRepository) List(ctx context.Context) ([]*Chunk, error) {
-	query := `
+	return r.list(ctx, `
 		SELECT chunk_hash, size
 		FROM chunks
 		ORDER BY chunk_hash
-	`
+	`)
+}
 
+// ListInUploadedBlobs returns the chunks that are stored in a blob whose
+// upload has completed (uploaded_ts set), ordered by chunk hash. These
+// are the only chunks a backup may safely deduplicate against: a chunk
+// recorded solely in a blob that was never uploaded refers to data that
+// is not in remote storage, so trusting it would silently drop that data
+// from later snapshots.
+func (r *ChunkRepository) ListInUploadedBlobs(ctx context.Context) ([]*Chunk, error) {
+	return r.list(ctx, `
+		SELECT DISTINCT c.chunk_hash, c.size
+		FROM chunks c
+		JOIN blob_chunks bc ON c.chunk_hash = bc.chunk_hash
+		JOIN blobs b ON bc.blob_id = b.id
+		WHERE b.uploaded_ts IS NOT NULL
+		ORDER BY c.chunk_hash
+	`)
+}
+
+// list runs a chunk-selecting query and scans the (chunk_hash, size) rows.
+func (r *ChunkRepository) list(ctx context.Context, query string) ([]*Chunk, error) {
 	rows, err := r.db.conn.QueryContext(ctx, query)
 	if err != nil {
 		return nil, fmt.Errorf("querying chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var chunks []*Chunk
 

internal/database/database.go

@@ -3,8 +3,10 @@
 //
 // Blobs in Vaultik are the final storage units uploaded to S3. Each blob is a
 // large (up to 10GB) file containing many compressed and encrypted chunks from
-// multiple source files. Blobs are content-addressed, meaning their filename
-// is derived from their SHA256 hash after compression and encryption.
+// multiple source files. Blobs are content-addressed: the filename in S3 is
+// hex(SHA256(SHA256(uncompressed blob contents))), computed from the chunk data
+// before compression and encryption (not from the stored bytes). See
+// blobgen.DoubleSHA256 and docs/REPOSTRUCTURE.md.
 //
 // Schema is managed via numbered SQL migrations embedded in the schema/
 // directory. Migration 000.sql bootstraps the schema_migrations tracking
@@ -17,6 +19,7 @@ import (
 	"embed"
 	"errors"
 	"fmt"
+	"net/url"
 	"os"
 	"path/filepath"
 	"sort"
@@ -39,6 +42,10 @@ var schemaFS embed.FS
 // table itself. It is applied before the normal migration loop.
 const bootstrapVersion = 0
 
+// busyTimeoutMs is how long a connection to the index waits for another
+// connection's lock before failing with "database is locked".
+const busyTimeoutMs = 10000
+
 // DB represents the Vaultik local index database connection.
 // It uses SQLite to track file metadata, content-defined chunks, and blob associations.
 // The database enables incremental backups by detecting changed files and
@@ -91,10 +98,27 @@ func ParseMigrationVersion(filename string) (int, error) {
 	return version, nil
 }
 
+// indexDSN returns the driver DSN that opens the database at path. The
+// driver runs each _pragma parameter on every connection it opens and drops
+// any parameter it does not know without an error, so a setting written in
+// another form silently does nothing. In WAL mode one connection can write
+// while others read; the busy timeout makes a connection wait for a lock
+// instead of failing at once.
+func indexDSN(path string) string {
+	return fmt.Sprintf(
+		"%s?_pragma=busy_timeout(%d)&_pragma=journal_mode(WAL)"+
+			"&_pragma=synchronous(NORMAL)&_pragma=foreign_keys(1)",
+		path, busyTimeoutMs)
+}
+
 // New creates a new database connection at the specified path.
-// It creates the schema if needed and configures SQLite with WAL mode for
-// better concurrency. SQLite handles crash recovery automatically when
-// opening a database with journal/WAL files present.
+// It creates the schema if needed. Every connection runs in WAL mode with
+// a busy timeout and foreign keys on (see indexDSN), so a read-only command
+// can read the index while a backup writes to it. Committed rows can sit in
+// the -wal file beside the database until a checkpoint, so a copy of the
+// database file alone may miss them.
+// SQLite handles crash recovery automatically when opening a database with
+// journal/WAL files present.
 // The path parameter can be a file path for persistent storage or ":memory:"
 // for an in-memory database (useful for testing).
 func New(ctx context.Context, path string) (*DB, error) {
@@ -107,11 +131,7 @@ func New(ctx context.Context, path string) (*DB, error) {
 	// First attempt with standard WAL mode
 	log.Debug("Attempting to open database with WAL mode", "path", path)
 
-	conn, err := sql.Open(
-		"sqlite",
-		path+"?_journal_mode=WAL&_synchronous=NORMAL&_busy_timeout=10000"+
-			"&_locking_mode=NORMAL&_foreign_keys=ON",
-	)
+	conn, err := sql.Open("sqlite", indexDSN(path))
 	if err == nil {
 		configureConnPool(conn)
 
@@ -131,8 +151,8 @@ func New(ctx context.Context, path string) (*DB, error) {
 		_ = conn.Close()
 	}
 
-	// If first attempt failed, try with TRUNCATE mode to clear any locks
-	return openWithRecovery(ctx, path)
+	// If the first attempt failed, try once more
+	return retryOpen(ctx, path)
 }
 
 // configureConnPool serializes all database access through one connection.
@@ -144,18 +164,12 @@ func configureConnPool(conn *sql.DB) {
 	conn.SetMaxIdleConns(1)
 }
 
-// finishOpen enables foreign keys, wraps the connection, and applies any
-// pending migrations. On migration failure the connection is closed.
+// finishOpen wraps the connection and applies any pending migrations. On
+// migration failure the connection is closed.
 func finishOpen(ctx context.Context, conn *sql.DB, path string) (*DB, error) {
-	// Enable foreign keys explicitly
-	_, err := conn.ExecContext(ctx, "PRAGMA foreign_keys = ON")
-	if err != nil {
-		log.Warn("Failed to enable foreign keys", "path", path, "error", err)
-	}
-
 	db := &DB{conn: conn, path: path}
 
-	err = applyMigrations(ctx, conn)
+	err := applyMigrations(ctx, conn)
 	if err != nil {
 		_ = conn.Close()
 
@@ -165,21 +179,15 @@ func finishOpen(ctx context.Context, conn *sql.DB, path string) (*DB, error) {
 	return db, nil
 }
 
-// openWithRecovery retries opening the database in TRUNCATE journal mode to
-// clear stale locks, then switches back to WAL mode.
-func openWithRecovery(ctx context.Context, path string) (*DB, error) {
-	log.Info(
-		"Database appears locked, attempting recovery with TRUNCATE mode",
-		"path", path,
-	)
+// retryOpen makes a second attempt to open the database, with the same
+// settings, after the first attempt failed, for example because another
+// process held a lock for longer than the busy timeout.
+func retryOpen(ctx context.Context, path string) (*DB, error) {
+	log.Info("Database appears locked, retrying open", "path", path)
 
-	conn, err := sql.Open(
-		"sqlite",
-		path+"?_journal_mode=TRUNCATE&_synchronous=NORMAL&_busy_timeout=10000"+
-			"&_foreign_keys=ON",
-	)
+	conn, err := sql.Open("sqlite", indexDSN(path))
 	if err != nil {
-		return nil, fmt.Errorf("opening database in recovery mode: %w", err)
+		return nil, fmt.Errorf("opening database on retry: %w", err)
 	}
 
 	configureConnPool(conn)
@@ -187,28 +195,18 @@ func openWithRecovery(ctx context.Context, path string) (*DB, error) {
 	err = conn.PingContext(ctx)
 	if err != nil {
 		log.Debug(
-			"Failed to ping database in recovery mode, closing",
+			"Failed to ping database on retry, closing",
 			"path", path, "error", err,
 		)
 
 		_ = conn.Close()
 
 		return nil, fmt.Errorf(
-			"database still locked after recovery attempt: %w",
+			"database still locked on retry: %w",
 			err,
 		)
 	}
 
-	log.Debug("Database opened in TRUNCATE mode", "path", path)
-
-	// Switch back to WAL mode
-	log.Debug("Switching database back to WAL mode", "path", path)
-
-	_, err = conn.ExecContext(ctx, "PRAGMA journal_mode=WAL")
-	if err != nil {
-		log.Warn("Failed to switch back to WAL mode", "path", path, "error", err)
-	}
-
 	db, err := finishOpen(ctx, conn, path)
 	if err != nil {
 		return nil, err
@@ -219,6 +217,135 @@ func openWithRecovery(ctx context.Context, path string) (*DB, error) {
 	return db, nil
 }
 
+// errUntrustedSnapshotSchema is returned when a downloaded snapshot
+// database carries schema objects the real schema never defines, or is
+// missing a table the restore and deep-verify queries read.
+var errUntrustedSnapshotSchema = errors.New(
+	"downloaded snapshot database has an untrusted schema")
+
+// snapshotReadOnlyDSN builds the driver DSN that opens a materialized
+// snapshot database file read-only. mode=ro opens the file read-only at
+// the OS level, query_only rejects any write the engine is asked to make,
+// and trusted_schema=OFF refuses to run application code named in the
+// schema. The file: URI form is required for the driver to honour the
+// mode parameter.
+func snapshotReadOnlyDSN(path string) string {
+	u := url.URL{
+		Scheme:   "file",
+		Path:     path,
+		RawQuery: "mode=ro&_pragma=query_only(true)&_pragma=trusted_schema(false)",
+	}
+
+	return u.String()
+}
+
+// OpenReadOnly opens an already-materialized SQLite file for read-only
+// querying of a snapshot database downloaded from the store, used by
+// restore and deep verify. Unlike New it never applies schema migrations
+// and never writes: the connection is opened read-only with query_only
+// and trusted_schema=OFF. It refuses any file whose schema carries a
+// trigger, view or virtual table, or lacks an expected table, so a forged
+// file cannot redefine what the restore queries return. The caller owns
+// the file and must remove it.
+func OpenReadOnly(ctx context.Context, path string) (*DB, error) {
+	conn, err := sql.Open("sqlite", snapshotReadOnlyDSN(path))
+	if err != nil {
+		return nil, fmt.Errorf("opening read-only database: %w", err)
+	}
+
+	configureConnPool(conn)
+
+	err = conn.PingContext(ctx)
+	if err != nil {
+		_ = conn.Close()
+
+		return nil, fmt.Errorf("opening read-only database: %w", err)
+	}
+
+	err = verifySnapshotSchema(ctx, conn)
+	if err != nil {
+		_ = conn.Close()
+
+		return nil, err
+	}
+
+	return &DB{conn: conn, path: path}, nil
+}
+
+// verifySnapshotSchema rejects a downloaded database whose schema is not
+// the plain table set the real schema defines. Any trigger, view or
+// virtual table, or a missing expected table, fails the open.
+func verifySnapshotSchema(ctx context.Context, conn *sql.DB) error {
+	// expectedSnapshotTables are the tables the restore and deep-verify
+	// queries read. A downloaded database missing any of them is not a
+	// genuine snapshot database and is refused.
+	expectedSnapshotTables := []string{
+		"blob_chunks",
+		"blobs",
+		"chunks",
+		"file_chunks",
+		"files",
+	}
+
+	rows, err := conn.QueryContext(
+		ctx, "SELECT type, name, sql FROM sqlite_master")
+	if err != nil {
+		return fmt.Errorf("reading snapshot schema: %w", err)
+	}
+
+	defer func() { _ = rows.Close() }()
+
+	present := make(map[string]struct{})
+
+	for rows.Next() {
+		var objType, name string
+
+		var objSQL sql.NullString
+
+		err = rows.Scan(&objType, &name, &objSQL)
+		if err != nil {
+			return fmt.Errorf("reading snapshot schema: %w", err)
+		}
+
+		switch objType {
+		case "trigger", "view":
+			return fmt.Errorf(
+				"%w: unexpected %s %q", errUntrustedSnapshotSchema, objType, name)
+		case "table":
+			if isVirtualTableSQL(objSQL.String) {
+				return fmt.Errorf(
+					"%w: unexpected virtual table %q",
+					errUntrustedSnapshotSchema, name)
+			}
+
+			present[name] = struct{}{}
+		}
+	}
+
+	err = rows.Err()
+	if err != nil {
+		return fmt.Errorf("reading snapshot schema: %w", err)
+	}
+
+	for _, table := range expectedSnapshotTables {
+		if _, ok := present[table]; !ok {
+			return fmt.Errorf(
+				"%w: missing table %q", errUntrustedSnapshotSchema, table)
+		}
+	}
+
+	return nil
+}
+
+// isVirtualTableSQL reports whether a sqlite_master row's SQL defines a
+// virtual table. Virtual tables are recorded with type 'table' but a
+// "CREATE VIRTUAL TABLE" definition and can run module code, so they are
+// refused alongside triggers and views.
+func isVirtualTableSQL(createSQL string) bool {
+	return strings.HasPrefix(
+		strings.ToUpper(strings.TrimSpace(createSQL)), "CREATE VIRTUAL TABLE")
+}
+
 // NewTestDB creates an in-memory SQLite database for testing purposes.
 // The database is automatically initialized with the schema and is ready
 // for use. Each call creates a new independent database instance.

internal/database/database_test.go

@@ -120,6 +120,89 @@ func TestDatabaseConcurrentAccess(t *testing.T) {
 	}
 }
 
+// TestNewSetsJournalModeAndBusyTimeout checks that the connection settings
+// New passes reach SQLite. The driver drops a setting it does not recognise
+// without an error, so only reading the value back shows it took effect.
+func TestNewSetsJournalModeAndBusyTimeout(t *testing.T) {
+	t.Parallel()
+
+	ctx := context.Background()
+
+	db, err := New(ctx, filepath.Join(t.TempDir(), "index.db"))
+	if err != nil {
+		t.Fatalf("failed to create database: %v", err)
+	}
+
+	defer func() { _ = db.Close() }()
+
+	var journalMode string
+
+	err = db.conn.QueryRowContext(ctx, "PRAGMA journal_mode").Scan(&journalMode)
+	if err != nil {
+		t.Fatalf("reading journal_mode: %v", err)
+	}
+
+	if journalMode != "wal" {
+		t.Errorf("journal_mode = %q, want %q", journalMode, "wal")
+	}
+
+	var busyTimeout int
+
+	err = db.conn.QueryRowContext(ctx, "PRAGMA busy_timeout").Scan(&busyTimeout)
+	if err != nil {
+		t.Fatalf("reading busy_timeout: %v", err)
+	}
+
+	if busyTimeout != busyTimeoutMs {
+		t.Errorf("busy_timeout = %d, want %d", busyTimeout, busyTimeoutMs)
+	}
+}
+
+// TestNewWriteSucceedsWhileAnotherHandleReads opens the same index twice,
+// as a read-only command does while a backup runs, and checks that a write
+// on one handle commits while the other is in the middle of a read.
+func TestNewWriteSucceedsWhileAnotherHandleReads(t *testing.T) {
+	t.Parallel()
+
+	ctx := context.Background()
+	dbPath := filepath.Join(t.TempDir(), "index.db")
+
+	reader, err := New(ctx, dbPath)
+	if err != nil {
+		t.Fatalf("failed to open reading handle: %v", err)
+	}
+
+	defer func() { _ = reader.Close() }()
+
+	writer, err := New(ctx, dbPath)
+	if err != nil {
+		t.Fatalf("failed to open writing handle: %v", err)
+	}
+
+	defer func() { _ = writer.Close() }()
+
+	// The read lock taken by the SELECT is held until the transaction ends.
+	readTx, err := reader.BeginTx(ctx, nil)
+	if err != nil {
+		t.Fatalf("beginning read transaction: %v", err)
+	}
+
+	defer func() { _ = readTx.Rollback() }()
+
+	var count int
+
+	err = readTx.QueryRowContext(ctx, "SELECT COUNT(*) FROM chunks").Scan(&count)
+	if err != nil {
+		t.Fatalf("reading chunks: %v", err)
+	}
+
+	_, err = writer.ExecWithLog(ctx,
+		"INSERT INTO chunks (chunk_hash, size) VALUES (?, ?)", "hash", 1024)
+	if err != nil {
+		t.Fatalf("write while another handle reads: %v", err)
+	}
+}
+
 func TestParseMigrationVersion(t *testing.T) {
 	t.Parallel()
 

internal/database/errors.go

@@ -1,12 +0,0 @@
-package database
-
-import (
-	"fmt"
-	"os"
-)
-
-// Fatalf prints an error message to stderr and exits with status 1
-func Fatalf(format string, args ...any) {
-	fmt.Fprintf(os.Stderr, "FATAL: "+format+"\n", args...)
-	os.Exit(1)
-}

internal/database/file_chunks.go

@@ -62,12 +62,7 @@ func (r *FileChunkRepository) GetByPath(
 		return nil, fmt.Errorf("querying file chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanFileChunks(rows)
 }
@@ -88,12 +83,7 @@ func (r *FileChunkRepository) GetByFileID(
 		return nil, fmt.Errorf("querying file chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanFileChunks(rows)
 }
@@ -117,12 +107,7 @@ func (r *FileChunkRepository) GetByPathTx(
 		return nil, fmt.Errorf("querying file chunks: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	fileChunks, err := r.scanFileChunks(rows)
 	LogSQL("GetByPathTx", "Complete", path, "count", len(fileChunks))
@@ -253,7 +238,7 @@ func (r *FileChunkRepository) CreateBatch(
 			args = append(args, fc.FileID.String(), fc.Idx, fc.ChunkHash.String())
 		}
 
-		query += querySb211.String() //nolint:gosec // G202: appends "?" placeholders only
+		query += querySb211.String()
 
 		query += " ON CONFLICT(file_id, idx) DO NOTHING"
 

internal/database/files.go

@@ -33,11 +33,13 @@ func (r *FileRepository) Create(ctx context.Context, tx *sql.Tx, file *File) err
 	}
 
 	query := `
-		INSERT INTO files (id, path, source_path, mtime, size, mode, uid, gid, link_target)
-		VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
+		INSERT INTO files
+			(id, path, source_path, mtime, mtime_nsec, size, mode, uid, gid, link_target)
+		VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
 		ON CONFLICT(path) DO UPDATE SET
 			source_path = excluded.source_path,
 			mtime = excluded.mtime,
+			mtime_nsec = excluded.mtime_nsec,
 			size = excluded.size,
 			mode = excluded.mode,
 			uid = excluded.uid,
@@ -54,16 +56,19 @@ func (r *FileRepository) Create(ctx context.Context, tx *sql.Tx, file *File) err
 	if tx != nil {
 		LogSQL("Execute", query,
 			file.ID.String(), file.Path.String(), file.SourcePath.String(),
-			file.MTime.Unix(), file.Size, file.Mode, file.UID, file.GID,
+			file.MTime.Unix(), file.MTime.Nanosecond(),
+			file.Size, file.Mode, file.UID, file.GID,
 			file.LinkTarget.String())
 		err = tx.QueryRowContext(ctx, query,
 			file.ID.String(), file.Path.String(), file.SourcePath.String(),
-			file.MTime.Unix(), file.Size, file.Mode, file.UID, file.GID,
+			file.MTime.Unix(), file.MTime.Nanosecond(),
+			file.Size, file.Mode, file.UID, file.GID,
 			file.LinkTarget.String()).Scan(&idStr)
 	} else {
 		err = r.db.QueryRowWithLog(ctx, query,
 			file.ID.String(), file.Path.String(), file.SourcePath.String(),
-			file.MTime.Unix(), file.Size, file.Mode, file.UID, file.GID,
+			file.MTime.Unix(), file.MTime.Nanosecond(),
+			file.Size, file.Mode, file.UID, file.GID,
 			file.LinkTarget.String()).Scan(&idStr)
 	}
 
@@ -84,7 +89,7 @@ func (r *FileRepository) Create(ctx context.Context, tx *sql.Tx, file *File) err
 // in the index.
 func (r *FileRepository) GetByPath(ctx context.Context, path string) (*File, error) {
 	query := `
-		SELECT id, path, source_path, mtime, size, mode, uid, gid, link_target
+		SELECT id, path, source_path, mtime, mtime_nsec, size, mode, uid, gid, link_target
 		FROM files
 		WHERE path = ?
 	`
@@ -104,7 +109,7 @@ func (r *FileRepository) GetByPath(ctx context.Context, path string) (*File, err
 // GetByID retrieves a file by its UUID
 func (r *FileRepository) GetByID(ctx context.Context, id types.FileID) (*File, error) {
 	query := `
-		SELECT id, path, source_path, mtime, size, mode, uid, gid, link_target
+		SELECT id, path, source_path, mtime, mtime_nsec, size, mode, uid, gid, link_target
 		FROM files
 		WHERE id = ?
 	`
@@ -127,7 +132,7 @@ func (r *FileRepository) GetByPathTx(
 	ctx context.Context, tx *sql.Tx, path string,
 ) (*File, error) {
 	query := `
-		SELECT id, path, source_path, mtime, size, mode, uid, gid, link_target
+		SELECT id, path, source_path, mtime, mtime_nsec, size, mode, uid, gid, link_target
 		FROM files
 		WHERE path = ?
 	`
@@ -158,23 +163,19 @@ func (r *FileRepository) ListModifiedSince(
 	ctx context.Context, since time.Time,
 ) ([]*File, error) {
 	query := `
-		SELECT id, path, source_path, mtime, size, mode, uid, gid, link_target
+		SELECT id, path, source_path, mtime, mtime_nsec, size, mode, uid, gid, link_target
 		FROM files
-		WHERE mtime >= ?
+		WHERE (mtime, mtime_nsec) >= (?, ?)
 		ORDER BY path
 	`
 
-	rows, err := r.db.conn.QueryContext(ctx, query, since.Unix())
+	rows, err := r.db.conn.QueryContext(ctx, query,
+		since.Unix(), since.Nanosecond())
 	if err != nil {
 		return nil, fmt.Errorf("querying files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var files []*File
 
@@ -228,29 +229,29 @@ func (r *FileRepository) DeleteByID(
 	return nil
 }
 
-// ListByPrefix returns all files whose path starts with prefix, ordered by
-// path.
-func (r *FileRepository) ListByPrefix(
-	ctx context.Context, prefix string,
+// ListUnderPath returns the file at path and every file beneath it,
+// ordered by path. Paths are compared case-sensitively, and a trailing
+// slash on path is ignored, so "/" lists every file.
+func (r *FileRepository) ListUnderPath(
+	ctx context.Context, path string,
 ) ([]*File, error) {
+	path = strings.TrimRight(path, "/")
+	dirPrefix := path + "/"
+
+	// LIKE would ignore ASCII case and treat _ and % in path as wildcards.
 	query := `
-		SELECT id, path, source_path, mtime, size, mode, uid, gid, link_target
+		SELECT id, path, source_path, mtime, mtime_nsec, size, mode, uid, gid, link_target
 		FROM files
-		WHERE path LIKE ? || '%'
+		WHERE path = ? OR substr(path, 1, length(?)) = ?
 		ORDER BY path
 	`
 
-	rows, err := r.db.conn.QueryContext(ctx, query, prefix)
+	rows, err := r.db.conn.QueryContext(ctx, query, path, dirPrefix, dirPrefix)
 	if err != nil {
 		return nil, fmt.Errorf("querying files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var files []*File
 
@@ -266,10 +267,94 @@ func (r *FileRepository) ListByPrefix(
 	return files, rows.Err()
 }
 
+// ListUnderPathInSnapshot returns the files that ListUnderPath returns
+// for path and that the snapshot lists, in no particular order.
+func (r *FileRepository) ListUnderPathInSnapshot(
+	ctx context.Context, path, snapshotID string,
+) ([]*File, error) {
+	path = strings.TrimRight(path, "/")
+	dirPrefix := path + "/"
+
+	query := `
+		SELECT f.id, f.path, f.source_path, f.mtime, f.mtime_nsec, f.size,
+			f.mode, f.uid, f.gid, f.link_target
+		FROM files f
+		JOIN snapshot_files sf ON sf.file_id = f.id
+		WHERE sf.snapshot_id = ?
+		  AND (f.path = ? OR substr(f.path, 1, length(?)) = ?)
+	`
+
+	rows, err := r.db.conn.QueryContext(ctx, query,
+		snapshotID, path, dirPrefix, dirPrefix)
+	if err != nil {
+		return nil, fmt.Errorf("querying files: %w", err)
+	}
+
+	defer func() { _ = rows.Close() }()
+
+	var files []*File
+
+	for rows.Next() {
+		file, err := r.scanFileRows(rows)
+		if err != nil {
+			return nil, fmt.Errorf("scanning file: %w", err)
+		}
+
+		files = append(files, file)
+	}
+
+	return files, rows.Err()
+}
+
+// ListIDsWithChunksNotInUploadedBlobs returns the IDs of the files whose
+// path starts with prefix and that list at least one chunk held by no
+// blob whose upload has completed (uploaded_ts set). A new snapshot
+// cannot reference such a chunk, so a backup must not treat the file as
+// unchanged even when its metadata matches the file on disk.
+func (r *FileRepository) ListIDsWithChunksNotInUploadedBlobs(
+	ctx context.Context, prefix string,
+) ([]types.FileID, error) {
+	query := `
+		SELECT DISTINCT f.id
+		FROM files f
+		JOIN file_chunks fc ON fc.file_id = f.id
+		WHERE f.path LIKE ? || '%'
+		  AND NOT EXISTS (
+			SELECT 1
+			FROM blob_chunks bc
+			JOIN blobs b ON bc.blob_id = b.id
+			WHERE bc.chunk_hash = fc.chunk_hash
+			  AND b.uploaded_ts IS NOT NULL
+		  )
+	`
+
+	rows, err := r.db.conn.QueryContext(ctx, query, prefix)
+	if err != nil {
+		return nil, fmt.Errorf("querying files: %w", err)
+	}
+
+	defer func() { _ = rows.Close() }()
+
+	var ids []types.FileID
+
+	for rows.Next() {
+		var id types.FileID
+
+		err := rows.Scan(&id)
+		if err != nil {
+			return nil, fmt.Errorf("scanning file ID: %w", err)
+		}
+
+		ids = append(ids, id)
+	}
+
+	return ids, rows.Err()
+}
+
 // ListAll returns all files in the database
 func (r *FileRepository) ListAll(ctx context.Context) ([]*File, error) {
 	query := `
-		SELECT id, path, source_path, mtime, size, mode, uid, gid, link_target
+		SELECT id, path, source_path, mtime, mtime_nsec, size, mode, uid, gid, link_target
 		FROM files
 		ORDER BY path
 	`
@@ -279,12 +364,7 @@ func (r *FileRepository) ListAll(ctx context.Context) ([]*File, error) {
 		return nil, fmt.Errorf("querying files: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	var files []*File
 
@@ -310,7 +390,7 @@ func (r *FileRepository) CreateBatch(
 	}
 
 	// Each files row binds this many SQL variables.
-	const fileCols = 9
+	const fileCols = 10
 
 	// Batch at 100 rows to be safe with SQLite's variable limit.
 	const batchSize = 100
@@ -321,7 +401,7 @@ func (r *FileRepository) CreateBatch(
 		batch := files[i:end]
 
 		query := `INSERT INTO files
-			(id, path, source_path, mtime, size, mode, uid, gid, link_target)
+			(id, path, source_path, mtime, mtime_nsec, size, mode, uid, gid, link_target)
 			VALUES `
 
 		args := make([]any, 0, len(batch)*fileCols)
@@ -333,19 +413,21 @@ func (r *FileRepository) CreateBatch(
 				querySb325.WriteString(", ")
 			}
 
-			querySb325.WriteString("(?, ?, ?, ?, ?, ?, ?, ?, ?)")
+			querySb325.WriteString("(?, ?, ?, ?, ?, ?, ?, ?, ?, ?)")
 
 			args = append(args,
 				f.ID.String(), f.Path.String(), f.SourcePath.String(),
-				f.MTime.Unix(), f.Size, f.Mode, f.UID, f.GID,
+				f.MTime.Unix(), f.MTime.Nanosecond(),
+				f.Size, f.Mode, f.UID, f.GID,
 				f.LinkTarget.String())
 		}
 
-		query += querySb325.String() //nolint:gosec // G202: appends "?" placeholders only
+		query += querySb325.String()
 
 		query += ` ON CONFLICT(path) DO UPDATE SET
 			source_path = excluded.source_path,
 			mtime = excluded.mtime,
+			mtime_nsec = excluded.mtime_nsec,
 			size = excluded.size,
 			mode = excluded.mode,
 			uid = excluded.uid,
@@ -405,7 +487,7 @@ func (r *FileRepository) scanFileFrom(row fileRowScanner) (*File, error) {
 	var (
 		file                          File
 		idStr, pathStr, sourcePathStr string
-		mtimeUnix                     int64
+		mtimeUnix, mtimeNsec          int64
 		linkTarget                    sql.NullString
 	)
 
@@ -414,6 +496,7 @@ func (r *FileRepository) scanFileFrom(row fileRowScanner) (*File, error) {
 		&pathStr,
 		&sourcePathStr,
 		&mtimeUnix,
+		&mtimeNsec,
 		&file.Size,
 		&file.Mode,
 		&file.UID,
@@ -432,7 +515,7 @@ func (r *FileRepository) scanFileFrom(row fileRowScanner) (*File, error) {
 	file.Path = types.FilePath(pathStr)
 	file.SourcePath = types.SourcePath(sourcePathStr)
 
-	file.MTime = time.Unix(mtimeUnix, 0).UTC()
+	file.MTime = time.Unix(mtimeUnix, mtimeNsec).UTC()
 	if linkTarget.Valid {
 		file.LinkTarget = types.FilePath(linkTarget.String)
 	}

internal/database/files_test.go

@@ -5,10 +5,12 @@ import (
 	"database/sql"
 	"errors"
 	"os"
+	"slices"
 	"testing"
 	"time"
 
 	"sneak.berlin/go/vaultik/internal/database"
+	"sneak.berlin/go/vaultik/internal/types"
 )
 
 // errTestRollback is the sentinel returned from transaction bodies to
@@ -134,6 +136,141 @@ func TestFileRepositoryListDelete(t *testing.T) {
 	}
 }
 
+func TestFileRepositoryListUnderPath(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx := context.Background()
+	repo := database.NewFileRepository(db)
+
+	const (
+		docDir  = "/home/u/doc"
+		docFile = "/home/u/doc/a.txt"
+	)
+
+	// In path order, so the root case can expect all of them as listed.
+	paths := []string{
+		"/home/u/50%/x.txt",
+		"/home/u/50percent/y.txt",
+		"/home/u/DOC/c.txt",
+		"/home/u/a_b/x.txt",
+		"/home/u/axb/y.txt",
+		docDir,
+		"/home/u/doc.txt.bak",
+		docFile,
+		"/home/u/doc/sub/b.txt",
+		"/home/u/doc2/b.txt",
+	}
+
+	for _, path := range paths {
+		err := repo.Create(ctx, nil, &database.File{
+			Path:  types.FilePath(path),
+			MTime: time.Now().Truncate(time.Second),
+			Mode:  0644,
+		})
+		if err != nil {
+			t.Fatalf("failed to create %s: %v", path, err)
+		}
+	}
+
+	docTree := []string{docDir, docFile, "/home/u/doc/sub/b.txt"}
+
+	tests := []struct {
+		name string
+		path string
+		want []string
+	}{
+		{"directory", docDir, docTree},
+		{"directory with trailing slash", docDir + "/", docTree},
+		{"directory differing only in case", "/home/u/DOC",
+			[]string{"/home/u/DOC/c.txt"}},
+		{"file", docFile, []string{docFile}},
+		{"underscore is literal", "/home/u/a_b",
+			[]string{"/home/u/a_b/x.txt"}},
+		{"percent is literal", "/home/u/50%",
+			[]string{"/home/u/50%/x.txt"}},
+		{"root", "/", paths},
+	}
+
+	for _, tt := range tests {
+		files, err := repo.ListUnderPath(ctx, tt.path)
+		if err != nil {
+			t.Fatalf("%s: failed to list files: %v", tt.name, err)
+		}
+
+		got := make([]string, 0, len(files))
+		for _, f := range files {
+			got = append(got, f.Path.String())
+		}
+
+		if !slices.Equal(got, tt.want) {
+			t.Errorf("%s: listing %q got %q, want %q",
+				tt.name, tt.path, got, tt.want)
+		}
+	}
+}
+
+// ListUnderPathInSnapshot leaves out files outside path and files the
+// snapshot does not list, though another snapshot lists them.
+func TestFileRepositoryListUnderPathInSnapshot(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx := context.Background()
+	files := database.NewFileRepository(db)
+	snapshots := database.NewSnapshotRepository(db)
+
+	listed := map[string][]string{
+		"one": {"/src/a.txt", "/src2/c.txt"},
+		"two": {"/src/b.txt"},
+	}
+
+	for snapshotID, paths := range listed {
+		err := snapshots.Create(ctx, nil, &database.Snapshot{
+			ID:        types.SnapshotID(snapshotID),
+			Hostname:  "host",
+			StartedAt: time.Now(),
+		})
+		if err != nil {
+			t.Fatalf("failed to create snapshot %s: %v", snapshotID, err)
+		}
+
+		for _, path := range paths {
+			err := files.Create(ctx, nil, &database.File{
+				Path:  types.FilePath(path),
+				MTime: time.Now().Truncate(time.Second),
+				Mode:  0644,
+			})
+			if err != nil {
+				t.Fatalf("failed to create %s: %v", path, err)
+			}
+
+			err = snapshots.AddFile(ctx, nil, snapshotID, path)
+			if err != nil {
+				t.Fatalf("failed to add %s to %s: %v", path, snapshotID, err)
+			}
+		}
+	}
+
+	found, err := files.ListUnderPathInSnapshot(ctx, "/src", "one")
+	if err != nil {
+		t.Fatalf("failed to list files: %v", err)
+	}
+
+	got := make([]string, 0, len(found))
+	for _, f := range found {
+		got = append(got, f.Path.String())
+	}
+
+	if !slices.Equal(got, []string{"/src/a.txt"}) {
+		t.Errorf("got %q, want only /src/a.txt", got)
+	}
+}
+
 func TestFileRepositorySymlink(t *testing.T) {
 	t.Parallel()
 
@@ -174,6 +311,115 @@ func TestFileRepositorySymlink(t *testing.T) {
 	}
 }
 
+// An mtime after 2262 or before 1678 does not fit in int64 nanoseconds
+// since the epoch, and must still come back from the database unchanged.
+func TestFileRepositoryMTimeOutsideInt64NanosecondRange(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx := context.Background()
+	repo := database.NewFileRepository(db)
+
+	mtimes := []time.Time{
+		time.Date(2300, time.January, 1, 0, 0, 0, 123456789, time.UTC),
+		time.Date(1601, time.January, 1, 0, 0, 0, 987654321, time.UTC),
+	}
+
+	for _, mtime := range mtimes {
+		created := &database.File{
+			Path:  types.FilePath("/created-" + mtime.Format(time.RFC3339Nano)),
+			MTime: mtime,
+		}
+
+		err := repo.Create(ctx, nil, created)
+		if err != nil {
+			t.Fatalf("failed to create file: %v", err)
+		}
+
+		batched := &database.File{
+			ID:    types.NewFileID(),
+			Path:  types.FilePath("/batched-" + mtime.Format(time.RFC3339Nano)),
+			MTime: mtime,
+		}
+
+		err = repo.CreateBatch(ctx, nil, []*database.File{batched})
+		if err != nil {
+			t.Fatalf("failed to batch create file: %v", err)
+		}
+
+		for _, path := range []types.FilePath{created.Path, batched.Path} {
+			retrieved, err := repo.GetByPath(ctx, path.String())
+			if err != nil {
+				t.Fatalf("failed to get file: %v", err)
+			}
+
+			if !retrieved.MTime.Equal(mtime) {
+				t.Errorf("%s: mtime got %v, want %v",
+					path, retrieved.MTime, mtime)
+			}
+		}
+	}
+}
+
+// A file already in the index and rewritten within the same second must get
+// its new nanoseconds stored, through both Create and CreateBatch.
+func TestFileRepositoryUpsertMTimeInSameSecond(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx := context.Background()
+	repo := database.NewFileRepository(db)
+
+	indexed := time.Date(2026, time.October, 7, 12, 0, 0, 100000000, time.UTC)
+	rewritten := indexed.Add(800 * time.Millisecond)
+
+	tests := []struct {
+		name   string
+		upsert func(file *database.File) error
+	}{
+		{"Create", func(file *database.File) error {
+			return repo.Create(ctx, nil, file)
+		}},
+		{"CreateBatch", func(file *database.File) error {
+			return repo.CreateBatch(ctx, nil, []*database.File{file})
+		}},
+	}
+
+	for _, tt := range tests {
+		file := &database.File{
+			ID:    types.NewFileID(),
+			Path:  types.FilePath("/" + tt.name),
+			MTime: indexed,
+		}
+
+		err := tt.upsert(file)
+		if err != nil {
+			t.Fatalf("%s: failed to create file: %v", tt.name, err)
+		}
+
+		file.MTime = rewritten
+
+		err = tt.upsert(file)
+		if err != nil {
+			t.Fatalf("%s: failed to update file: %v", tt.name, err)
+		}
+
+		retrieved, err := repo.GetByPath(ctx, file.Path.String())
+		if err != nil {
+			t.Fatalf("%s: failed to get file: %v", tt.name, err)
+		}
+
+		if !retrieved.MTime.Equal(rewritten) {
+			t.Errorf("%s: mtime got %v, want %v",
+				tt.name, retrieved.MTime, rewritten)
+		}
+	}
+}
+
 func TestFileRepositoryTransaction(t *testing.T) {
 	t.Parallel()
 

internal/database/models.go

@@ -14,8 +14,7 @@ type File struct {
 	ID   types.FileID   // UUID primary key
 	Path types.FilePath // Absolute path of the file
 
-	// SourcePath is the source directory this file came from (used for
-	// restore path stripping).
+	// SourcePath is the source directory this file came from.
 	SourcePath types.SourcePath
 	MTime      time.Time
 	Size       int64
@@ -51,15 +50,15 @@ type Chunk struct {
 // Blob represents a blob record in the database.
 // A blob is Vaultik's final storage unit - a large file (up to 10GB) containing
 // many compressed and encrypted chunks from multiple source files.
-// Blobs are content-addressed, meaning their filename in S3 is derived from
-// the SHA256 hash of their compressed and encrypted content.
-// The blob creation process is: chunks are accumulated -> compressed with zstd
-// -> encrypted with age -> hashed -> uploaded to S3 with the hash as filename.
+// Blobs are content-addressed: the filename in S3 is
+// hex(SHA256(SHA256(uncompressed blob contents))), computed from the chunk data
+// before compression and encryption (not from the stored bytes). See
+// blobgen.DoubleSHA256 and docs/REPOSTRUCTURE.md.
 type Blob struct {
 	ID types.BlobID // UUID assigned when blob creation starts
 
-	// Hash is the SHA256 of the final compressed+encrypted content
-	// (empty until finalized).
+	// Hash is hex(SHA256(SHA256(uncompressed blob contents)))
+	// (empty until finalized); see the type comment above.
 	Hash             types.BlobHash
 	CreatedTS        time.Time  // When blob creation started
 	FinishedTS       *time.Time // When blob was finalized (nil if still packing)
@@ -99,8 +98,8 @@ type Snapshot struct {
 	StartedAt          time.Time
 	CompletedAt        *time.Time // nil if still in progress
 	FileCount          int64
-	ChunkCount         int64
-	BlobCount          int64
+	ChunkCount         int64 // Chunks this snapshot stored that were not stored before
+	BlobCount          int64 // Blobs this snapshot created
 	TotalSize          int64 // Total size of all referenced files
 
 	// BlobSize is the total size of all referenced blobs (compressed and

internal/database/module.go

@@ -2,6 +2,7 @@ package database
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"os"
 	"path/filepath"
@@ -15,6 +16,44 @@ import (
 // the index describes the backed-up file tree and must stay private.
 const indexDirPerm = 0o700
 
+// indexFilePerm restricts the index file to the owning user; it lists every
+// backed-up path and chunk hash and must stay private.
+const indexFilePerm = 0o600
+
+// ensureIndexFileMode makes the index file owner-only before the SQLite
+// driver opens it: it creates the file 0600 if absent, or chmods an existing
+// one to 0600. Doing this first matters because SQLite creates its -wal and
+// -shm side files with the mode of the main database file, so a private main
+// file yields private side files. The driver treats a zero-byte file as an
+// empty database, so pre-creating it here is safe.
+func ensureIndexFileMode(path string) error {
+	info, err := os.Stat(path)
+
+	switch {
+	case err == nil:
+		if info.Mode().Perm() == indexFilePerm {
+			return nil
+		}
+
+		err = os.Chmod(path, indexFilePerm)
+		if err != nil {
+			return fmt.Errorf("restricting index file permissions: %w", err)
+		}
+
+		return nil
+	case errors.Is(err, os.ErrNotExist):
+		//nolint:gosec // G304: the index path is operator-configured by design
+		f, err := os.OpenFile(path, os.O_CREATE|os.O_WRONLY, indexFilePerm)
+		if err != nil {
+			return fmt.Errorf("creating index file: %w", err)
+		}
+
+		return f.Close()
+	default:
+		return fmt.Errorf("checking index file: %w", err)
+	}
+}
+
 // Module provides database dependencies
 //
 //nolint:gochecknoglobals // fx module definitions are package globals by convention
@@ -34,6 +73,11 @@ func provideDatabase(lc fx.Lifecycle, cfg *config.Config) (*DB, error) {
 		return nil, fmt.Errorf("creating index directory: %w", err)
 	}
 
+	err = ensureIndexFileMode(cfg.IndexPath)
+	if err != nil {
+		return nil, err
+	}
+
 	db, err := New(context.Background(), cfg.IndexPath)
 	if err != nil {
 		return nil, fmt.Errorf("opening database: %w", err)

internal/database/module_internal_test.go

@@ -0,0 +1,100 @@
+package database
+
+import (
+	"os"
+	"path/filepath"
+	"syscall"
+	"testing"
+
+	"go.uber.org/fx/fxtest"
+	"sneak.berlin/go/vaultik/internal/config"
+)
+
+// TestProvideDatabaseFreshIndexMode verifies that provideDatabase creates a
+// missing index file owner-only (0600), even under a lenient 022 umask that
+// would otherwise leave a freshly created file world-readable.
+//
+//nolint:paralleltest // syscall.Umask is process-global; parallel tests would clash
+func TestProvideDatabaseFreshIndexMode(t *testing.T) {
+	restore := syscall.Umask(0o022)
+	defer syscall.Umask(restore)
+
+	indexPath := filepath.Join(t.TempDir(), "index.sqlite")
+
+	openIndex(t, indexPath)
+	assertPerm(t, indexPath, 0o600)
+}
+
+// TestProvideDatabaseExistingIndexMode verifies that provideDatabase tightens
+// an existing world-readable index (0644) in a group/other-readable directory
+// down to owner-only (0600).
+//
+//nolint:paralleltest // syscall.Umask is process-global; parallel tests would clash
+func TestProvideDatabaseExistingIndexMode(t *testing.T) {
+	restore := syscall.Umask(0o022)
+	defer syscall.Umask(restore)
+
+	dir := filepath.Join(t.TempDir(), "data")
+
+	//nolint:gosec // G301: the test intentionally uses a 0755 directory
+	err := os.MkdirAll(dir, 0o755)
+	if err != nil {
+		t.Fatalf("creating index directory: %v", err)
+	}
+
+	//nolint:gosec // G302: the test intentionally uses a 0755 directory
+	err = os.Chmod(dir, 0o755)
+	if err != nil {
+		t.Fatalf("relaxing index directory permissions: %v", err)
+	}
+
+	indexPath := filepath.Join(dir, "index.sqlite")
+
+	//nolint:gosec // G306: the test intentionally starts from a 0644 index
+	err = os.WriteFile(indexPath, nil, 0o644)
+	if err != nil {
+		t.Fatalf("creating pre-existing index: %v", err)
+	}
+
+	//nolint:gosec // G302: the test intentionally starts from a 0644 index
+	err = os.Chmod(indexPath, 0o644)
+	if err != nil {
+		t.Fatalf("relaxing pre-existing index permissions: %v", err)
+	}
+
+	openIndex(t, indexPath)
+	assertPerm(t, indexPath, 0o600)
+}
+
+// openIndex runs provideDatabase against indexPath and closes the resulting
+// database before returning.
+func openIndex(t *testing.T, indexPath string) {
+	t.Helper()
+
+	cfg := &config.Config{IndexPath: indexPath}
+
+	db, err := provideDatabase(fxtest.NewLifecycle(t), cfg)
+	if err != nil {
+		t.Fatalf("provideDatabase: %v", err)
+	}
+
+	err = db.Close()
+	if err != nil {
+		t.Fatalf("closing database: %v", err)
+	}
+}
+
+// assertPerm fails the test unless path has exactly the given permission bits.
+func assertPerm(t *testing.T, path string, want os.FileMode) {
+	t.Helper()
+
+	info, err := os.Stat(path)
+	if err != nil {
+		t.Fatalf("stat %s: %v", path, err)
+	}
+
+	got := info.Mode().Perm()
+	if got != want {
+		t.Fatalf("permissions of %s = %#o, want %#o", path, got, want)
+	}
+}

internal/database/read_only_test.go

@@ -0,0 +1,145 @@
+//nolint:testpackage // exercises unexported read-only open internals
+package database
+
+import (
+	"context"
+	"database/sql"
+	"errors"
+	"path/filepath"
+	"testing"
+)
+
+// genuineSnapshotDB writes a real snapshot database (the full schema
+// applied) to a fresh file and returns its path.
+func genuineSnapshotDB(t *testing.T) string {
+	t.Helper()
+
+	path := filepath.Join(t.TempDir(), "snapshot.db")
+
+	db, err := New(context.Background(), path)
+	if err != nil {
+		t.Fatalf("creating snapshot database: %v", err)
+	}
+
+	err = db.Close()
+	if err != nil {
+		t.Fatalf("closing snapshot database: %v", err)
+	}
+
+	return path
+}
+
+// forgedDB creates an empty database file and runs the given statements
+// against it read-write, so a test can plant schema objects the real
+// schema never defines.
+func forgedDB(t *testing.T, stmts ...string) string {
+	t.Helper()
+
+	path := filepath.Join(t.TempDir(), "forged.db")
+
+	db, err := sql.Open("sqlite", path)
+	if err != nil {
+		t.Fatalf("opening forged database: %v", err)
+	}
+
+	for _, stmt := range stmts {
+		_, err = db.ExecContext(context.Background(), stmt)
+		if err != nil {
+			t.Fatalf("executing %q: %v", stmt, err)
+		}
+	}
+
+	err = db.Close()
+	if err != nil {
+		t.Fatalf("closing forged database: %v", err)
+	}
+
+	return path
+}
+
+func TestOpenReadOnlyAcceptsGenuineSnapshot(t *testing.T) {
+	t.Parallel()
+
+	db, err := OpenReadOnly(context.Background(), genuineSnapshotDB(t))
+	if err != nil {
+		t.Fatalf("OpenReadOnly refused a genuine snapshot database: %v", err)
+	}
+
+	t.Cleanup(func() { _ = db.Close() })
+}
+
+func TestOpenReadOnlyRefusesWrites(t *testing.T) {
+	t.Parallel()
+
+	db, err := OpenReadOnly(context.Background(), genuineSnapshotDB(t))
+	if err != nil {
+		t.Fatalf("OpenReadOnly: %v", err)
+	}
+
+	t.Cleanup(func() { _ = db.Close() })
+
+	// A schema write depends on no table columns, so the only reason it
+	// can fail is that the database is open read-only.
+	_, err = db.Conn().ExecContext(context.Background(),
+		"CREATE TABLE probe_readonly (x)")
+	if err == nil {
+		t.Fatal("expected a write to a read-only snapshot database to fail")
+	}
+}
+
+func TestOpenReadOnlyRejectsView(t *testing.T) {
+	t.Parallel()
+
+	path := forgedDB(t, "CREATE VIEW files AS SELECT 1 AS path")
+
+	_, err := OpenReadOnly(context.Background(), path)
+	if !errors.Is(err, errUntrustedSnapshotSchema) {
+		t.Fatalf("expected a view named files to be refused, got %v", err)
+	}
+}
+
+func TestOpenReadOnlyRejectsTrigger(t *testing.T) {
+	t.Parallel()
+
+	path := forgedDB(t,
+		"CREATE TABLE files (path TEXT)",
+		"CREATE TRIGGER t AFTER INSERT ON files BEGIN SELECT 1; END")
+
+	_, err := OpenReadOnly(context.Background(), path)
+	if !errors.Is(err, errUntrustedSnapshotSchema) {
+		t.Fatalf("expected a trigger to be refused, got %v", err)
+	}
+}
+
+func TestOpenReadOnlyRejectsMissingTable(t *testing.T) {
+	t.Parallel()
+
+	// Only one of the expected tables is present.
+	path := forgedDB(t, "CREATE TABLE files (path TEXT)")
+
+	_, err := OpenReadOnly(context.Background(), path)
+	if !errors.Is(err, errUntrustedSnapshotSchema) {
+		t.Fatalf("expected a missing expected table to be refused, got %v", err)
+	}
+}
+
+func TestIsVirtualTableSQL(t *testing.T) {
+	t.Parallel()
+
+	cases := []struct {
+		sql  string
+		want bool
+	}{
+		{"CREATE VIRTUAL TABLE t USING fts5(x)", true},
+		{"  create virtual table t using fts5(x)", true},
+		{"CREATE TABLE t (x)", false},
+		{"CREATE VIEW t AS SELECT 1", false},
+		{"", false},
+	}
+
+	for _, c := range cases {
+		if got := isVirtualTableSQL(c.sql); got != c.want {
+			t.Errorf("isVirtualTableSQL(%q) = %v, want %v", c.sql, got, c.want)
+		}
+	}
+}

internal/database/repositories.go

@@ -3,6 +3,7 @@ package database
 import (
 	"context"
 	"database/sql"
+	"errors"
 	"fmt"
 )
 
@@ -60,22 +61,24 @@ func (r *Repositories) WithTx(ctx context.Context, fn TxFunc) error {
 
 	defer func() {
 		if p := recover(); p != nil {
-			rollbackErr := tx.Rollback()
-			if rollbackErr != nil {
-				Fatalf("failed to rollback transaction: %v", rollbackErr)
-			}
+			// The panic is the failure to report, so a rollback error
+			// is dropped.
+			_ = tx.Rollback()
 
 			panic(p)
-		} else if err != nil {
-			rollbackErr := tx.Rollback()
-			if rollbackErr != nil {
-				Fatalf("failed to rollback transaction: %v", rollbackErr)
-			}
 		}
 	}()
 
 	err = fn(ctx, tx)
 	if err != nil {
+		// Once ctx is cancelled, database/sql rolls the transaction
+		// back itself and Rollback returns sql.ErrTxDone.
+		rollbackErr := tx.Rollback()
+		if rollbackErr != nil && !errors.Is(rollbackErr, sql.ErrTxDone) {
+			return fmt.Errorf("%w; rolling back transaction: %w",
+				err, rollbackErr)
+		}
+
 		return err
 	}
 
@@ -103,22 +106,24 @@ func (r *Repositories) WithReadTx(ctx context.Context, fn TxFunc) error {
 
 	defer func() {
 		if p := recover(); p != nil {
-			rollbackErr := tx.Rollback()
-			if rollbackErr != nil {
-				Fatalf("failed to rollback transaction: %v", rollbackErr)
-			}
+			// The panic is the failure to report, so a rollback error
+			// is dropped.
+			_ = tx.Rollback()
 
 			panic(p)
-		} else if err != nil {
-			rollbackErr := tx.Rollback()
-			if rollbackErr != nil {
-				Fatalf("failed to rollback transaction: %v", rollbackErr)
-			}
 		}
 	}()
 
 	err = fn(ctx, tx)
 	if err != nil {
+		// Once ctx is cancelled, database/sql rolls the transaction
+		// back itself and Rollback returns sql.ErrTxDone.
+		rollbackErr := tx.Rollback()
+		if rollbackErr != nil && !errors.Is(rollbackErr, sql.ErrTxDone) {
+			return fmt.Errorf("%w; rolling back transaction: %w",
+				err, rollbackErr)
+		}
+
 		return err
 	}
 

internal/database/repositories_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"database/sql"
 	"errors"
+	"strings"
 	"testing"
 	"time"
 
@@ -235,6 +236,125 @@ func TestRepositoriesTransactionRollback(t *testing.T) {
 	}
 }
 
+// When the context is cancelled during the body, database/sql rolls the
+// transaction back itself, so WithTx's own rollback finds it already
+// done. WithTx must return the body's error alone, without that rollback.
+func TestRepositoriesTransactionContextCancelled(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	repos := database.NewRepositories(db)
+
+	err := repos.WithTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+		file := &database.File{
+			Path:  "/test/cancelled_file.txt",
+			MTime: time.Now().Truncate(time.Second),
+			Size:  1024,
+			Mode:  0644,
+			UID:   1000,
+			GID:   1000,
+		}
+
+		err := repos.Files.Create(ctx, tx, file)
+		if err != nil {
+			return err
+		}
+
+		cancel()
+
+		// database/sql rolls back from a goroutine of its own; wait
+		// until it has, so the test does not depend on that race.
+		for {
+			_, err = tx.ExecContext(context.WithoutCancel(ctx), "SELECT 1")
+			if errors.Is(err, sql.ErrTxDone) {
+				return ctx.Err()
+			}
+
+			time.Sleep(time.Millisecond)
+		}
+	})
+	if !errors.Is(err, context.Canceled) {
+		t.Fatalf("expected context.Canceled, got: %v", err)
+	}
+
+	if errors.Is(err, sql.ErrTxDone) {
+		t.Fatalf("the already-done rollback was reported: %v", err)
+	}
+
+	file, err := repos.Files.GetByPath(
+		context.Background(), "/test/cancelled_file.txt",
+	)
+	if err != nil {
+		t.Fatalf("error checking for file: %v", err)
+	}
+
+	if file != nil {
+		t.Error("file should not exist after the cancelled transaction")
+	}
+}
+
+// A body that ends the transaction itself with ROLLBACK makes WithTx's
+// own rollback fail. WithTx must return that failure together with the
+// body's error.
+func TestRepositoriesTransactionRollbackFails(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx := context.Background()
+	repos := database.NewRepositories(db)
+
+	err := repos.WithTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+		_, err := tx.ExecContext(ctx, "ROLLBACK")
+		if err != nil {
+			return err
+		}
+
+		return errIntentionalRollback
+	})
+	if !errors.Is(err, errIntentionalRollback) {
+		t.Fatalf("expected the body's error, got: %v", err)
+	}
+
+	if !strings.Contains(err.Error(), "rolling back transaction") {
+		t.Errorf("expected the failed rollback in the error, got: %v", err)
+	}
+}
+
+// When the body ends the transaction itself with ROLLBACK and then
+// panics, WithTx's own rollback fails; the panic must still reach the
+// caller.
+func TestRepositoriesTransactionPanicRollbackFails(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx := context.Background()
+	repos := database.NewRepositories(db)
+
+	defer func() {
+		if recover() == nil {
+			t.Error("expected the body's panic to reach the caller")
+		}
+	}()
+
+	_ = repos.WithTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+		_, err := tx.ExecContext(ctx, "ROLLBACK")
+		if err != nil {
+			return err
+		}
+
+		panic(errIntentionalRollback)
+	})
+}
+
 func TestRepositoriesReadTransaction(t *testing.T) {
 	t.Parallel()
 
@@ -291,3 +411,41 @@ func TestRepositoriesReadTransaction(t *testing.T) {
 		t.Error("expected to retrieve file in read transaction")
 	}
 }
+
+// When the context is cancelled during a read transaction's body,
+// database/sql rolls the transaction back itself, so WithReadTx's own
+// rollback finds it already done. WithReadTx must return the body's error
+// alone, without that rollback.
+func TestRepositoriesReadTransactionContextCancelled(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+
+	repos := database.NewRepositories(db)
+
+	err := repos.WithReadTx(ctx, func(ctx context.Context, tx *sql.Tx) error {
+		cancel()
+
+		// database/sql rolls back from a goroutine of its own; wait
+		// until it has, so the test does not depend on that race.
+		for {
+			_, err := tx.ExecContext(context.WithoutCancel(ctx), "SELECT 1")
+			if errors.Is(err, sql.ErrTxDone) {
+				return ctx.Err()
+			}
+
+			time.Sleep(time.Millisecond)
+		}
+	})
+	if !errors.Is(err, context.Canceled) {
+		t.Fatalf("expected context.Canceled, got: %v", err)
+	}
+
+	if errors.Is(err, sql.ErrTxDone) {
+		t.Fatalf("the already-done rollback was reported: %v", err)
+	}
+}

internal/database/repository_comprehensive_test.go

@@ -824,7 +824,7 @@ func TestTransactionIsolation(t *testing.T) {
 	}
 
 	// Verify the file was not created (transaction rolled back)
-	files, err := repos.Files.ListByPrefix(ctx, "/tx-test")
+	files, err := repos.Files.ListUnderPath(ctx, "/tx-test.txt")
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -916,7 +916,7 @@ func TestConcurrentOrphanedCleanup(t *testing.T) {
 	}
 
 	// Verify correct files were deleted
-	files, err := repos.Files.ListByPrefix(ctx, "/concurrent-")
+	files, err := repos.Files.ListAll(ctx)
 	if err != nil {
 		t.Fatal(err)
 	}

internal/database/repository_debug_test.go

@@ -147,7 +147,7 @@ func TestOrphanedFileCleanupDebug(t *testing.T) {
 	t.Logf("Files count after cleanup: %d", count)
 
 	// List remaining files
-	files, err := repos.Files.ListByPrefix(ctx, "/")
+	files, err := repos.Files.ListUnderPath(ctx, "/")
 	if err != nil {
 		t.Fatal(err)
 	}

internal/database/repository_edge_cases_test.go

@@ -3,6 +3,7 @@ package database
 
 import (
 	"context"
+	"database/sql"
 	"fmt"
 	"strings"
 	"testing"
@@ -367,7 +368,7 @@ func verifyBlobNullUploadTS(
 }
 
 // createLargeDatasetFiles creates fileCount files and adds every other
-// one to the snapshot.
+// one to the snapshot, in one transaction as a backup writes them.
 func createLargeDatasetFiles(
 	t *testing.T,
 	repos *Repositories,
@@ -376,9 +377,10 @@ func createLargeDatasetFiles(
 ) {
 	t.Helper()
 
-	ctx := context.Background()
 	start := time.Now()
 
+	err := repos.WithTx(context.Background(),
+		func(ctx context.Context, tx *sql.Tx) error {
 			for i := range fileCount {
 				file := &File{
 					Path:  types.FilePath(fmt.Sprintf("/large/file%05d.txt", i)),
@@ -389,19 +391,25 @@ func createLargeDatasetFiles(
 					GID:   uint32(1000 + (i % 10)),
 				}
 
-		err := repos.Files.Create(ctx, nil, file)
+				err := repos.Files.Create(ctx, tx, file)
 				if err != nil {
-			t.Fatalf("failed to create file %d: %v", i, err)
+					return fmt.Errorf("creating file %d: %w", i, err)
 				}
 
 				// Add half to snapshot
 				if i%2 == 0 {
-			err = repos.Snapshots.AddFileByID(ctx, nil, snapshotID, file.ID)
+					err = repos.Snapshots.AddFileByID(ctx, tx, snapshotID, file.ID)
+					if err != nil {
+						return err
+					}
+				}
+			}
+
+			return nil
+		})
 	if err != nil {
 		t.Fatal(err)
 	}
-		}
-	}
 
 	t.Logf("Created %d files in %v", fileCount, time.Since(start))
 }
@@ -442,12 +450,12 @@ func TestLargeDatasets(t *testing.T) {
 		createLargeDatasetFiles(t, repos, snapshot.ID.String(), fileCount)
 	})
 
-	// Test ListByPrefix performance
+	// Test ListUnderPath performance
 	//nolint:paralleltest // phases share one database and are order-dependent
-	t.Run("list by prefix performance", func(t *testing.T) {
+	t.Run("list under path performance", func(t *testing.T) {
 		start := time.Now()
 
-		files, err := repos.Files.ListByPrefix(ctx, "/large/")
+		files, err := repos.Files.ListUnderPath(ctx, "/large/")
 		if err != nil {
 			t.Fatal(err)
 		}
@@ -472,7 +480,7 @@ func TestLargeDatasets(t *testing.T) {
 		t.Logf("Cleaned up orphaned files in %v", time.Since(start))
 
 		// Verify correct number remain
-		files, err := repos.Files.ListByPrefix(ctx, "/large/")
+		files, err := repos.Files.ListUnderPath(ctx, "/large/")
 		if err != nil {
 			t.Fatal(err)
 		}
@@ -606,8 +614,7 @@ func TestTimezoneHandling(t *testing.T) {
 		t.Skip("timezone not available")
 	}
 
-	// Use Truncate to remove sub-second precision since we store as Unix timestamps
-	nyTime := time.Now().In(loc).Truncate(time.Second)
+	nyTime := time.Now().In(loc)
 	file := &File{
 		Path:  "/timezone-test.txt",
 		MTime: nyTime,

internal/database/schema/001.sql

@@ -5,8 +5,9 @@
 CREATE TABLE IF NOT EXISTS files (
     id TEXT PRIMARY KEY,  -- UUID
     path TEXT NOT NULL UNIQUE,
-    source_path TEXT NOT NULL DEFAULT '',  -- The source directory this file came from (for restore path stripping)
-    mtime INTEGER NOT NULL,
+    source_path TEXT NOT NULL DEFAULT '',  -- The source directory this file came from
+    mtime INTEGER NOT NULL,  -- whole seconds since the Unix epoch
+    mtime_nsec INTEGER NOT NULL,  -- nanoseconds within that second, 0 to 999999999
     size INTEGER NOT NULL,
     mode INTEGER NOT NULL,
     uid INTEGER NOT NULL,
@@ -85,8 +86,10 @@ CREATE TABLE IF NOT EXISTS snapshots (
     completed_at INTEGER,
     file_count INTEGER NOT NULL DEFAULT 0,
     chunk_count INTEGER NOT NULL DEFAULT 0,
+    new_chunk_size INTEGER NOT NULL DEFAULT 0,
     blob_count INTEGER NOT NULL DEFAULT 0,
     total_size INTEGER NOT NULL DEFAULT 0,
+    uncompressed_chunk_size INTEGER NOT NULL DEFAULT 0,
     blob_size INTEGER NOT NULL DEFAULT 0,
     blob_uncompressed_size INTEGER NOT NULL DEFAULT 0,
     compression_ratio REAL NOT NULL DEFAULT 1.0,
@@ -120,6 +123,16 @@ CREATE TABLE IF NOT EXISTS snapshot_blobs (
 -- Index for efficient blob lookups (used in orphan detection)
 CREATE INDEX IF NOT EXISTS idx_snapshot_blobs_blob_id ON snapshot_blobs(blob_id);
 
+-- Snapshot mount points table: the directories at or below a configured
+-- path that the snapshot's backup found on a different filesystem from
+-- their parent directory, as a mount point is, and found an entry in
+CREATE TABLE IF NOT EXISTS snapshot_mount_points (
+    snapshot_id TEXT NOT NULL,
+    path TEXT NOT NULL,
+    PRIMARY KEY (snapshot_id, path),
+    FOREIGN KEY (snapshot_id) REFERENCES snapshots(id) ON DELETE CASCADE
+);
+
 -- Uploads table: tracks blob upload metrics
 CREATE TABLE IF NOT EXISTS uploads (
     blob_hash TEXT PRIMARY KEY,

internal/database/snapshots.go

@@ -11,8 +11,21 @@ import (
 	"sneak.berlin/go/vaultik/internal/types"
 )
 
+// Sentinel errors for the single-snapshot invariant that an exported
+// per-snapshot metadata database must satisfy.
+var (
+	// ErrNoSnapshotInDatabase means the metadata database has no snapshot
+	// row at all.
+	ErrNoSnapshotInDatabase = errors.New("database contains no snapshot")
+	// ErrMultipleSnapshotsInDatabase means the metadata database holds
+	// more than the single snapshot an export is supposed to contain.
+	ErrMultipleSnapshotsInDatabase = errors.New(
+		"database contains more than one snapshot")
+)
+
 // SnapshotRepository provides access to the snapshots table and its
-// snapshot_files / snapshot_blobs association tables.
+// snapshot_files / snapshot_blobs association tables, and to the mount
+// points each snapshot's backup found (snapshot_mount_points).
 type SnapshotRepository struct {
 	db *DB
 }
@@ -108,14 +121,20 @@ func (r *SnapshotRepository) UpdateCounts(
 	return nil
 }
 
-// UpdateExtendedStats updates extended statistics for a snapshot
+// UpdateExtendedStats updates extended statistics for a snapshot. It also
+// records the total plaintext size of the distinct chunks the snapshot's
+// files reference. Only the snapshot's own backup can take that total: a
+// file's row is shared by every snapshot that lists it, and a later
+// backup of a changed file rewrites the row with the new chunks.
 func (r *SnapshotRepository) UpdateExtendedStats(
 	ctx context.Context,
 	tx *sql.Tx,
 	snapshotID string,
 	blobUncompressedSize int64,
 	compressionLevel int,
+	uploadBytes int64,
 	uploadDurationMs int64,
+	newChunkSize int64,
 ) error {
 	compressionRatio, err := r.extendedCompressionRatio(
 		ctx, tx, snapshotID, blobUncompressedSize,
@@ -129,19 +148,30 @@ func (r *SnapshotRepository) UpdateExtendedStats(
 		SET blob_uncompressed_size = ?,
 		    compression_ratio = ?,
 		    compression_level = ?,
-		    upload_bytes = blob_size,
-		    upload_duration_ms = ?
+		    upload_bytes = ?,
+		    upload_duration_ms = ?,
+		    new_chunk_size = ?,
+		    uncompressed_chunk_size = (
+		        SELECT COALESCE(SUM(c.size), 0)
+		        FROM (
+		            SELECT DISTINCT fc.chunk_hash
+		            FROM snapshot_files sf
+		            JOIN file_chunks fc ON sf.file_id = fc.file_id
+		            WHERE sf.snapshot_id = snapshots.id
+		        ) sc
+		        JOIN chunks c ON sc.chunk_hash = c.chunk_hash
+		    )
 		WHERE id = ?
 	`
 
 	if tx != nil {
 		_, err = tx.ExecContext(ctx, query,
 			blobUncompressedSize, compressionRatio, compressionLevel,
-			uploadDurationMs, snapshotID)
+			uploadBytes, uploadDurationMs, newChunkSize, snapshotID)
 	} else {
 		_, err = r.db.ExecWithLog(ctx, query,
 			blobUncompressedSize, compressionRatio, compressionLevel,
-			uploadDurationMs, snapshotID)
+			uploadBytes, uploadDurationMs, newChunkSize, snapshotID)
 	}
 
 	if err != nil {
@@ -206,6 +236,43 @@ func (r *SnapshotRepository) GetByID(
 	return &snapshot, nil
 }
 
+// GetOnlySnapshot returns the sole snapshot in an exported per-snapshot
+// metadata database. The backup path writes each snapshot's database with
+// exactly one snapshot row (see cleanSnapshotDB), so restore and deep
+// verify expect exactly one. Zero rows return ErrNoSnapshotInDatabase and
+// more than one returns ErrMultipleSnapshotsInDatabase; callers treat
+// either as a failed identity check on the downloaded database.
+func (r *SnapshotRepository) GetOnlySnapshot(ctx context.Context) (*Snapshot, error) {
+	query := `
+		SELECT id, hostname, vaultik_version, vaultik_git_revision,
+			started_at, completed_at, file_count, chunk_count, blob_count,
+			total_size, blob_size, compression_ratio
+		FROM snapshots
+		LIMIT 2
+	`
+
+	rows, err := r.db.conn.QueryContext(ctx, query)
+	if err != nil {
+		return nil, fmt.Errorf("querying snapshots: %w", err)
+	}
+
+	defer func() { _ = rows.Close() }()
+
+	snapshots, err := r.scanSnapshotRows(rows)
+	if err != nil {
+		return nil, err
+	}
+
+	switch len(snapshots) {
+	case 1:
+		return snapshots[0], nil
+	case 0:
+		return nil, ErrNoSnapshotInDatabase
+	default:
+		return nil, ErrMultipleSnapshotsInDatabase
+	}
+}
+
 // ListRecent returns up to limit snapshots, most recently started first.
 func (r *SnapshotRepository) ListRecent(
 	ctx context.Context, limit int,
@@ -224,12 +291,7 @@ func (r *SnapshotRepository) ListRecent(
 		return nil, fmt.Errorf("querying snapshots: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanSnapshotRows(rows)
 }
@@ -341,7 +403,7 @@ func (r *SnapshotRepository) AddFilesByIDBatch(
 			args = append(args, snapshotID, fileID.String())
 		}
 
-		query += querySb312.String() //nolint:gosec // G202: appends "?" placeholders only
+		query += querySb312.String()
 
 		var err error
 		if tx != nil {
@@ -439,32 +501,57 @@ func (r *SnapshotRepository) GetBlobHashes(
 		ORDER BY sb.blob_hash
 	`
 
-	rows, err := r.db.conn.QueryContext(ctx, query, snapshotID)
+	blobs, err := r.queryStrings(ctx, query, snapshotID)
 	if err != nil {
 		return nil, fmt.Errorf("querying blob hashes: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
+	return blobs, nil
+}
+
+// AddMountPoint records dir as a directory that the backup of a snapshot
+// found on a different filesystem from its parent directory, as a mount
+// point is, using tx when non-nil.
+func (r *SnapshotRepository) AddMountPoint(
+	ctx context.Context, tx *sql.Tx, snapshotID string, dir string,
+) error {
+	query := `
+		INSERT OR IGNORE INTO snapshot_mount_points (snapshot_id, path)
+		VALUES (?, ?)
+	`
+
+	var err error
+	if tx != nil {
+		_, err = tx.ExecContext(ctx, query, snapshotID, dir)
+	} else {
+		_, err = r.db.ExecWithLog(ctx, query, snapshotID, dir)
+	}
+
 	if err != nil {
-			Fatalf("failed to close rows: %v", err)
+		return fmt.Errorf("adding mount point to snapshot: %w", err)
 	}
-	}()
 
-	var blobs []string
+	return nil
+}
 
-	for rows.Next() {
-		var blobHash string
+// GetMountPoints returns the directories AddMountPoint recorded for a
+// snapshot, sorted.
+func (r *SnapshotRepository) GetMountPoints(
+	ctx context.Context, snapshotID string,
+) ([]string, error) {
+	query := `
+		SELECT path
+		FROM snapshot_mount_points
+		WHERE snapshot_id = ?
+		ORDER BY path
+	`
 
-		err := rows.Scan(&blobHash)
+	dirs, err := r.queryStrings(ctx, query, snapshotID)
 	if err != nil {
-			return nil, fmt.Errorf("scanning blob hash: %w", err)
+		return nil, fmt.Errorf("querying mount points: %w", err)
 	}
 
-		blobs = append(blobs, blobHash)
-	}
-
-	return blobs, rows.Err()
+	return dirs, nil
 }
 
 // GetSnapshotTotalCompressedSize returns the total compressed size of all
@@ -489,71 +576,66 @@ func (r *SnapshotRepository) GetSnapshotTotalCompressedSize(
 	return totalSize, nil
 }
 
-// GetSnapshotUncompressedChunkSize returns the sum of plaintext sizes of all unique
-// chunks referenced by a snapshot (via snapshot_files → file_chunks → chunks).
+// GetSnapshotBlobSizes returns the total compressed and uncompressed sizes
+// of all blobs referenced by a snapshot.
+func (r *SnapshotRepository) GetSnapshotBlobSizes(
+	ctx context.Context, snapshotID string,
+) (int64, int64, error) {
+	query := `
+		SELECT COALESCE(SUM(b.compressed_size), 0),
+			COALESCE(SUM(b.uncompressed_size), 0)
+		FROM snapshot_blobs sb
+		JOIN blobs b ON sb.blob_hash = b.blob_hash
+		WHERE sb.snapshot_id = ?
+	`
+
+	var compressed, uncompressed int64
+
+	err := r.db.conn.QueryRowContext(ctx, query, snapshotID).Scan(
+		&compressed, &uncompressed)
+	if err != nil {
+		return 0, 0, fmt.Errorf("querying snapshot blob sizes: %w", err)
+	}
+
+	return compressed, uncompressed, nil
+}
+
+// GetSnapshotUncompressedChunkSize returns the total plaintext size of the
+// distinct chunks the snapshot's files reference, as the snapshot's backup
+// recorded it (see UpdateExtendedStats).
 func (r *SnapshotRepository) GetSnapshotUncompressedChunkSize(
 	ctx context.Context, snapshotID string,
 ) (int64, error) {
-	query := `
-		SELECT COALESCE(SUM(c.size), 0)
-		FROM (
-			SELECT DISTINCT fc.chunk_hash
-			FROM snapshot_files sf
-			JOIN file_chunks fc ON sf.file_id = fc.file_id
-			WHERE sf.snapshot_id = ?
-		) sc
-		JOIN chunks c ON sc.chunk_hash = c.chunk_hash
-	`
+	query := `SELECT uncompressed_chunk_size FROM snapshots WHERE id = ?`
 
-	var totalSize int64
+	var uncompressedChunkSize int64
 
-	err := r.db.conn.QueryRowContext(ctx, query, snapshotID).Scan(&totalSize)
+	err := r.db.conn.QueryRowContext(ctx, query, snapshotID).Scan(&uncompressedChunkSize)
 	if err != nil {
 		return 0, fmt.Errorf("querying uncompressed chunk size: %w", err)
 	}
 
-	return totalSize, nil
+	return uncompressedChunkSize, nil
 }
 
-// GetSnapshotNewChunkSize returns the sum of plaintext sizes of chunks that are
-// referenced by this snapshot but not by any earlier completed snapshot known to
-// the local database. The result is the marginal uncompressed data this snapshot
-// added to the dedup pool — i.e., the delta from prior snapshots.
+// GetSnapshotNewChunkSize returns the total plaintext size of the chunks the
+// snapshot's backup stored that were not stored before, as the backup
+// recorded it. It is not derived from the snapshot's files: a file's row is
+// shared by every snapshot that lists it and is rewritten when the file
+// changes, so it shows only the file's latest chunks.
 func (r *SnapshotRepository) GetSnapshotNewChunkSize(
 	ctx context.Context, snapshotID string,
 ) (int64, error) {
-	query := `
-		WITH this_snap_chunks AS (
-			SELECT DISTINCT fc.chunk_hash
-			FROM snapshot_files sf
-			JOIN file_chunks fc ON sf.file_id = fc.file_id
-			WHERE sf.snapshot_id = ?
-		),
-		prior_chunks AS (
-			SELECT DISTINCT fc.chunk_hash
-			FROM snapshots s
-			JOIN snapshot_files sf ON sf.snapshot_id = s.id
-			JOIN file_chunks fc ON fc.file_id = sf.file_id
-			WHERE s.completed_at IS NOT NULL
-			  AND s.id != ?
-			  AND s.started_at < (SELECT started_at FROM snapshots WHERE id = ?)
-		)
-		SELECT COALESCE(SUM(c.size), 0)
-		FROM chunks c
-		JOIN this_snap_chunks t ON c.chunk_hash = t.chunk_hash
-		WHERE c.chunk_hash NOT IN (SELECT chunk_hash FROM prior_chunks)
-	`
+	query := `SELECT new_chunk_size FROM snapshots WHERE id = ?`
 
-	var totalSize int64
+	var newChunkSize int64
 
-	err := r.db.conn.QueryRowContext(
-		ctx, query, snapshotID, snapshotID, snapshotID,
-	).Scan(&totalSize)
+	err := r.db.conn.QueryRowContext(ctx, query, snapshotID).Scan(&newChunkSize)
 	if err != nil {
 		return 0, fmt.Errorf("querying new chunk size: %w", err)
 	}
 
-	return totalSize, nil
+	return newChunkSize, nil
 }
 
 // GetIncompleteSnapshots returns all snapshots that haven't been completed
@@ -574,12 +656,7 @@ func (r *SnapshotRepository) GetIncompleteSnapshots(
 		return nil, fmt.Errorf("querying incomplete snapshots: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	return r.scanSnapshotRows(rows)
 }
@@ -602,12 +679,7 @@ func (r *SnapshotRepository) GetIncompleteByHostname(
 		return nil, fmt.Errorf("querying incomplete snapshots: %w", err)
 	}
 
-	defer func() {
-		err := rows.Close()
-		if err != nil {
-			Fatalf("failed to close rows: %v", err)
-		}
-	}()
+	defer func() { _ = rows.Close() }()
 
 	// Same column set as every other multi-row snapshot query, so the
 	// shared scanner applies — including its timestamp normalization.
@@ -668,6 +740,34 @@ func (r *SnapshotRepository) DeleteSnapshotUploads(
 	return nil
 }
 
+// queryStrings runs query, which selects one text column, and returns
+// its values in the order of the rows.
+func (r *SnapshotRepository) queryStrings(
+	ctx context.Context, query string, args ...any,
+) ([]string, error) {
+	rows, err := r.db.conn.QueryContext(ctx, query, args...)
+	if err != nil {
+		return nil, err
+	}
+
+	defer func() { _ = rows.Close() }()
+
+	var values []string
+
+	for rows.Next() {
+		var value string
+
+		err := rows.Scan(&value)
+		if err != nil {
+			return nil, err
+		}
+
+		values = append(values, value)
+	}
+
+	return values, rows.Err()
+}
+
 // extendedCompressionRatio computes the compression ratio for a snapshot
 // from its stored blob_size and the given uncompressed size. Returns 1.0
 // when the uncompressed size is zero.

internal/database/snapshots_test.go

@@ -4,6 +4,7 @@ import (
 	"context"
 	"fmt"
 	"math"
+	"slices"
 	"testing"
 	"time"
 
@@ -145,6 +146,65 @@ func TestSnapshotRepositoryUpdateCounts(t *testing.T) {
 	}
 }
 
+// GetSnapshotBlobSizes totals the blobs the snapshot references, and only
+// those.
+func TestSnapshotRepositoryGetSnapshotBlobSizes(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx := context.Background()
+	repos := database.NewRepositories(db)
+
+	snapshot := &database.Snapshot{
+		ID:             "2024-01-03T12:00:00Z",
+		Hostname:       testHostname,
+		VaultikVersion: testVersion,
+		StartedAt:      time.Now().Truncate(time.Second),
+	}
+
+	err := repos.Snapshots.Create(ctx, nil, snapshot)
+	if err != nil {
+		t.Fatalf("failed to create snapshot: %v", err)
+	}
+
+	blobs := []*database.Blob{
+		{Hash: "referenced-1", CompressedSize: 10, UncompressedSize: 100},
+		{Hash: "referenced-2", CompressedSize: 20, UncompressedSize: 200},
+		{Hash: "unreferenced", CompressedSize: 40, UncompressedSize: 400},
+	}
+
+	for _, blob := range blobs {
+		blob.ID = types.NewBlobID()
+		blob.CreatedTS = time.Now().Truncate(time.Second)
+
+		err = repos.Blobs.Create(ctx, nil, blob)
+		if err != nil {
+			t.Fatalf("failed to create blob %s: %v", blob.Hash, err)
+		}
+	}
+
+	for _, blob := range blobs[:2] {
+		err = repos.Snapshots.AddBlob(ctx, nil, snapshot.ID.String(),
+			blob.ID, blob.Hash)
+		if err != nil {
+			t.Fatalf("failed to add blob %s to snapshot: %v", blob.Hash, err)
+		}
+	}
+
+	compressed, uncompressed, err := repos.Snapshots.GetSnapshotBlobSizes(
+		ctx, snapshot.ID.String())
+	if err != nil {
+		t.Fatalf("failed to get snapshot blob sizes: %v", err)
+	}
+
+	if compressed != 30 || uncompressed != 300 {
+		t.Errorf("blob sizes: got %d and %d, want 30 and 300",
+			compressed, uncompressed)
+	}
+}
+
 func TestSnapshotRepositoryListRecent(t *testing.T) {
 	t.Parallel()
 
@@ -363,3 +423,52 @@ func TestSnapshotRepositoryDuplicate(t *testing.T) {
 		t.Error("expected error for duplicate snapshot")
 	}
 }
+
+// Deleting a snapshot from the local index deletes the mount points its
+// backup recorded, and no other snapshot's. See
+// https://git.eeqj.de/sneak/vaultik/issues/395.
+func TestSnapshotRepositoryDeleteRemovesMountPoints(t *testing.T) {
+	t.Parallel()
+
+	db, cleanup := setupTestDB(t)
+	defer cleanup()
+
+	ctx := context.Background()
+	repo := database.NewSnapshotRepository(db)
+
+	for _, id := range []types.SnapshotID{"kept", "deleted"} {
+		err := repo.Create(ctx, nil, &database.Snapshot{
+			ID:             id,
+			Hostname:       testHostname,
+			VaultikVersion: testVersion,
+			StartedAt:      time.Now().Truncate(time.Second),
+		})
+		if err != nil {
+			t.Fatalf("failed to create snapshot: %v", err)
+		}
+
+		err = repo.AddMountPoint(ctx, nil, id.String(), "/data/nas")
+		if err != nil {
+			t.Fatalf("failed to add mount point: %v", err)
+		}
+	}
+
+	err := repo.Delete(ctx, "deleted")
+	if err != nil {
+		t.Fatalf("failed to delete snapshot: %v", err)
+	}
+
+	for id, want := range map[string][]string{
+		"kept":    {"/data/nas"},
+		"deleted": nil,
+	} {
+		got, err := repo.GetMountPoints(ctx, id)
+		if err != nil {
+			t.Fatalf("failed to read mount points: %v", err)
+		}
+
+		if !slices.Equal(got, want) {
+			t.Errorf("mount points of %s: got %v, want %v", id, got, want)
+		}
+	}
+}

internal/database/uploads.go

@@ -158,19 +158,3 @@ type UploadStats struct {
 	MinDurationMs int64
 	MaxDurationMs int64
 }
-
-// GetCountBySnapshot returns the count of uploads for a specific snapshot
-func (r *UploadRepository) GetCountBySnapshot(
-	ctx context.Context, snapshotID string,
-) (int64, error) {
-	query := `SELECT COUNT(*) FROM uploads WHERE snapshot_id = ?`
-
-	var count int64
-
-	err := r.conn.QueryRowContext(ctx, query, snapshotID).Scan(&count)
-	if err != nil {
-		return 0, err
-	}
-
-	return count, nil
-}

internal/globals/globals.go

@@ -3,20 +3,34 @@
 package globals
 
 import (
+	"regexp"
+	"strings"
 	"time"
 )
 
 // Appname is the application name, populated from main().
 var Appname = "vaultik" //nolint:gochecknoglobals // set via -ldflags at build time
 
+// DevVersion is the version a binary reports when it was built without
+// git metadata: script/version emits it outside a git checkout, and an
+// unstamped `go build` keeps it. goreleaser's snapshot template stamps
+// it followed by "-" and the commit it was built from. It is
+// deliberately not a number.
+const DevVersion = "dev"
+
+// Unknown is what Commit and CommitDate hold when the build did not
+// stamp them, and the version script/docker and script/cibuild stamp
+// when the host has no git checkout.
+const Unknown = "unknown"
+
 // Version is the application version, populated from main().
-var Version = "dev" //nolint:gochecknoglobals // set via -ldflags at build time
+var Version = DevVersion //nolint:gochecknoglobals // set via -ldflags at build time
 
 // Commit is the git commit hash, populated from main().
-var Commit = "unknown" //nolint:gochecknoglobals // set via -ldflags at build time
+var Commit = Unknown //nolint:gochecknoglobals // set via -ldflags at build time
 
 // CommitDate is the ISO-8601 date of the commit, populated from main().
-var CommitDate = "unknown" //nolint:gochecknoglobals // set via -ldflags at build time
+var CommitDate = Unknown //nolint:gochecknoglobals // set via -ldflags at build time
 
 // Author identifies the upstream author of vaultik.
 const Author = "Jeffrey Paul <sneak@sneak.berlin>"
@@ -24,6 +38,9 @@ const Author = "Jeffrey Paul <sneak@sneak.berlin>"
 // Homepage is the canonical URL for vaultik.
 const Homepage = "https://sneak.berlin/go/vaultik"
 
+// ReleasesURL is where tagged release artifacts are published.
+const ReleasesURL = "https://git.eeqj.de/sneak/vaultik/releases"
+
 // License is the SPDX identifier for the project license.
 const License = "MIT"
 
@@ -47,6 +64,31 @@ func New() (*Globals, error) {
 	}, nil
 }
 
+// IsDevVersion reports whether v names a development build rather than
+// a release. "dev" and goreleaser's snapshot "dev-<sha>" count, and so
+// does what `git describe --tags --always --dirty` gives a make or
+// docker build of an untagged commit: the bare short commit, or
+// tag-N-gHASH on a commit after a tag. Any version ending in "-dirty"
+// counts, a modified checkout of a tag ("v1.0.0-dirty") included.
+// A plain tag such as "v1.0.0" or "1.0.0" is a release.
+//
+// The empty string counts too. Nothing that knows its version reports
+// no version, so an empty Version means the stamping failed, and the
+// safe reading of "we could not establish that this is a release" is
+// that it is not one. The Makefile refuses to build at all in that
+// case; this is the second line of defence, for a binary linked by
+// something other than the Makefile. Unknown counts for the same
+// reason.
+func IsDevVersion(v string) bool {
+	if v == "" || v == Unknown || v == DevVersion ||
+		strings.HasPrefix(v, DevVersion+"-") ||
+		strings.HasSuffix(v, "-dirty") {
+		return true
+	}
+
+	return regexp.MustCompile(`^(.+-[0-9]+-g)?[0-9a-f]+$`).MatchString(v)
+}
+
 // shortCommitLen is the number of commit-hash characters ShortCommit keeps.
 const shortCommitLen = 12