24 KiB
Workflow
One issue per unit of work, one branch and one PR per issue:
- ensure a tracked issue exists with a definition of done
- branch from
next(never frommain) - do the work; open a PR based on
next(never onmain) - pass an independent review, then the manager squash-merges into
next - push; nothing stays local-only
next is the branch for the next milestone and must stay green and
mergeable to main without notice. One next -> main PR accumulates
the milestone; releases are cut from main separately.
Issue branches do NOT touch this file — the manager maintains it on
next. Every branch editing TODO.md conflicts with every other
(#112).
Status
1.0.0 is open, with work remaining. The milestone (https://git.eeqj.de/sneak/webhooker/milestone/9) is the authoritative list, and the only place to read a count or a state of play from. This file records where the project is, not what is in flight: a sentence whose truth depends on a branch being unmerged is wrong the moment it merges, and this file has been wrong that way before.
The durability defect that held the tag has landed
(#256, commit 8d64259).
Every SQLite handle opens with WAL journaling and a busy timeout, a
bookkeeping write that fails leaves its delivery in a recoverable
state rather than a lying one, and recovery skips a delivery that
already has a successful result row. Final pre-tag verification
exercised it and confirmed it holds. Whatever the milestone still
shows open is what remains before v1.0.0.
Delivery is at-least-once by design, not by accident: a send whose result row does not land is attempted again, so a receiver can see a duplicate. That is deliberate — the alternative is a silent lost delivery — and the README says so under Rationale. It is not a defect to re-file.
One caveat on reading a green check: a docs-only commit deliberately
replays from the layer cache
(#119), so a green status on
such a commit evidences a replay rather than an executed run. A code
commit invalidates the COPY layer and genuinely executes.
Next Step
Clear the rest of the open 1.0.0 milestone
(https://git.eeqj.de/sneak/webhooker/milestone/9) and tag v1.0.0.
Merging next into main is a separate act from tagging and waits on
neither of those: next is kept mergeable at all times, which is the
point of the branch.
Completed Steps
- 2026-08-24 Bind the plaintext HTTP listener deliberately, via
BIND_ADDRESSdefaulting to127.0.0.1, and document the reverse-proxy deployment. A hostname, an empty value or a value carrying a port is a startup error, and theDockerfilesets0.0.0.0because a loopback bind inside a container is unreachable (#268). The same commit removed the shutdown race:httpServeris built in the constructor rather than assigned from the serving goroutine, which orders the write before every fx hook and rules out the nil dereference a SIGTERM arriving first would have caused, andsentryEnabledis anatomic.Bool(#226) - 2026-08-24 Remove inbound request signature verification. The
entrypoint UUID is the authentication secret, so the per-entrypoint
shared secret, the
internal/signaturepackage, the receiver check, the model fields and the forms are all gone. This reverses the feature that landed earlier in the same milestone (#67, #279) - 2026-08-24 Stamp the build version into the binary and render it in
the UI footer.
script/versionis the single source —$VERSION, elsegit describe --tags --always --dirty, elseunknown— so amake buildbinary and amake dockerimage from one checkout report the same thing, and nothing in it varies between two builds of the same commit, which the release gate's byte-identical assertion would catch (#253) - 2026-08-24 Derive cookie
Secureand CSRF strictness from the request transport rather than fromWEBHOOKER_ENVIRONMENT. Behind a real TLS proxy with the environment left at itsdevdefault, the session cookie silently lostSecurewhile the CSRF cookie on the same response kept it.X-Forwarded-Protois now matched case-insensitively on its first comma-separated element, soHTTPSandhttps, httpno longer fall to the relaxed CSRF path (#269) - 2026-08-24 Roll back a failed webhook deletion instead of committing it. A failing delete committed whatever had already succeeded, hard-deleted the per-webhook event database anyway, and redirected as though it had worked — orphaned config plus permanently destroyed history, reported as success. All three delete positions now roll back with the event database intact (#262)
- 2026-08-24 Name a deleted target on its historical deliveries, marked
(deleted), rather than leaving the event log unable to say where a delivery went. A deleted target's credentials stay masked exactly as a live one's, and it cannot become deliverable again through the receiver, resubmit, replay, the edit form or the toggle (#211) - 2026-08-24 Bound both request-controlled
/metricslabel dimensions, so the unauthenticated receiver is no longer a memory-exhaustion vector:handlercarries the chi route pattern, andmethodfolds anything chi cannot route onto a single(unmatched)sentinel. Both were reproduced before the fix — 300 random method tokens took the series count from 106 to 7,631, and path flooding reached 62,532 — and a label audit across a live scrape found no third unbounded dimension (#254, #261) - 2026-08-24 Validate
max_retrieson both target forms.abc,2.7and-5silently became 0 — fire-and-forget — including on the edit path, where it destroyed a working value, and999999999stored verbatim. The ceiling of 20 is themaxboth templates already declared (#221) - 2026-08-24 Resubmit a stored event as a new undelivered event, so a backend under development can be tested against real captured traffic. Per-delivery replay cannot serve that: it re-sends one finished delivery to its own original target, and a target created for a dev backend has no prior delivery to replay. Resubmit re-injects the stored event at the top of the receiver path and fans it out to whatever targets are active now (#250)
- 2026-08-20 Take an exclusive lock on
DATA_DIRat startup, so two instances on one directory cannot both deliver (#201) - 2026-08-20 Shut down the app when the HTTP listener fails. The
OnStarthook returned as soon as the serving goroutine was spawned, so a failed listen left fx reporting RUNNING and a live process with nothing bound — invisible to systemd and Docker restart policies (#200) - 2026-08-20 Stop target credentials leaking into the per-webhook event databases (#206), log SQL with placeholders rather than bound values (#207), and fail loudly on half-set metrics auth credentials (#205)
- 2026-08-20 Read queue depths with
Find, notScan.Scanswaps GORM's own trace recorder in for the logging adapter, and that recorder does not implementgorm.ParamsFilter, so those statements logged their bound values interpolated and bypassed the suppression above. The two units gated green against anextthat lacked the other, andnextwent red when both landed (#234) - 2026-08-20 Render per-attempt delivery detail in the event log (#202) and add replay of a terminally failed delivery (#203)
- 2026-08-20 Expose delivery metrics on
/metrics(#209) and document the backup, restore and upgrade procedures (#210) - 2026-08-20 Add a
webhooker resetpwsubcommand and a bootstrap banner. The admin bootstrap password was printed once among roughly 45 fx lines, and underdocker run -dwent to container logs subject to rotation; there was no reset path at all, so recovery meant hand-deleting the users row, documented nowhere. The password is read from stdin or generated, never from argv where/procwould publish it (#208) - 2026-08-20 Add
ALLOWED_EGRESS_CIDRS, an allowlist-only escape hatch for the SSRF guard, so a self-hosted proxy can forward into the operator's own network. The guard's always-blocked set cannot be reopened by configuration (#204) - 2026-08-20 Harden operator-set target headers, which were carried unsafely across a redirect (#233)
- 2026-08-20 Add a target edit form with headers and timeout fields (#127)
- 2026-08-18 Raise
script/test's per-package timeout from 30s to 90s, matching the org-wide backstop.go testapplies-timeoutper package, andinternal/handlershad grown past the old budget: a cache-defeated build failed outright atGOMAXPROCS=4, and every run under deliberate host load breached 30s. The measurement table lives in the script (#194) - 2026-08-18 Re-sync
REPO_POLICIES.mdfromprompts. The local copy was stale and still mandated a 20s test target with a 30s timeout, which the org replaced with a 60s cap and a 90s backstop. A synced copy is not a source; reading it as one nearly produced a PR againstpromptsproposing a change already merged there (#196) - 2026-08-18 Report handler panics through the logger and answer 500.
chi v1.5.5's
Recovererscans for apanic(0xframe the runtime no longer emits, then indexespkg[-1:], so it panicked inside its own stack printer before writing a byte: the recovery never ran, the client got a dropped connection instead of a 500, and the original panic was lost. A local middleware replaces it, bounded byMaxPanicLogLineBytes(#187) - 2026-08-18 Route GORM's logger through
slogand bound it. Everygorm.Openleftlogger.Defaultin place atWarnwithIgnoreRecordNotFoundErrorfalse, so every record-not-found printed the fully interpolated SQL to stdout — including the client-chosen path on/webhook/{uuid}and the submitted username on the login form, at no level the operator set and outsideinternal/loggerentirely. Three call sites, not the two the issue named (#178) - 2026-08-18 Bound every
slogline against client-chosen text. Eight sites reachable unauthenticated, found by reading everyslogcall in the tree rather than only the one reported; the budget moved to a sharedinternal/logfieldso no second truncation exists.DEBUGbeing off by default is not a bound and is not treated as one (#176) - 2026-08-18 Stop a slow host turning a login-guard test into a
segfault. A non-fatal
asserton an acquire result was dereferenced on the next line, so one timing miss killed the wholeinternal/middlewarebinary and reddened CI for unrelated PRs. The fix also removed a real production race —acquirecould shed a request with a slot standing free, because Go picks uniformly among readyselectcases (#186) - 2026-08-18 Send the chi route pattern to Sentry rather than the
concrete path. The receiver's path carries the entrypoint capability
token, so every Sentry event from
/webhook/{uuid}shipped a live credential to a third party. RequestData,QueryString,CookiesandEnvare dropped and headers reduced to an allowlist (#179) - 2026-08-18 Read form fields from the POST body only.
r.FormValuemerges the query string, so a login could be driven by URL parameters — putting the password somewhere that lands in access logs, proxy logs and browser history (#160) - 2026-08-18 Verify login credentials before spending rate-limit budget, so a flood of wrong passwords cannot lock out the account it is guessing at. The manager took this decision rather than stall the queue; it is flagged on the issue for reversal (#150)
- 2026-08-18 Run all linting in Docker via
Dockerfile.lint. Host lint was wrong in both directions from version skew and shared caches.script/lintasserts the summary line, because--no-cache-filtersilently ignores a stage name it does not match — the flag that makes the gate meaningful fails open (#109) - 2026-08-18 Serve an event's full stored body over HTTP. The list query truncates for rendering, and that truncated value was the only way to read a body, so the full payload was unreachable (#157)
- 2026-08-18 Bound the access log line against client-chosen text.
internal/logfieldbudgets by encoded bytes, not runes, so a handler's JSON escaping cannot multiply a field past its allowance (#146) - 2026-08-18 Mark superseded CI commits
failurerather thanskipped. A skipped run rolls up green, so a commit that was never tested reported success (#152) - 2026-08-18 Set
fx.StopTimeoutinside the container stop grace, so shutdown hooks are bounded by a deadline the orchestrator will actually honour rather than being killed mid-flush (#134) - 2026-08-17 Bucket IPv6 rate-limit keys by
/64. A single allocation hands out 2^64 addresses, so per-address keying let one client mint unlimited buckets. Manager decision, recorded on the issue (#125) - 2026-08-17 Correct release-blocking README and startup-warning inaccuracies, including claims about behaviour the code does not have (#151)
- 2026-08-17 Fetch and verify Alpine.js at build time against
static/vendor.sha256instead of committing the minified blob, so the dependency is pinned by hash rather than by trust (#145) - 2026-08-17 Bound the event log's rendered bodies in the query itself, so a large stored payload cannot be read into memory just to be truncated for display (#135)
- 2026-08-17 Mask the
httptarget's destination URL in the UI: it can carry a bearer credential in its path or query, and was rendered verbatim. Manager decision to mask unconditionally (#115) - 2026-08-14 Bound shutdown hooks by their stop context, so a hook that hangs cannot hold the process past its grace period (#102)
- 2026-08-14 Render templates via a buffer rather than the
ResponseWriter, so a template error part-way through cannot commit a 200 and then fail — the response is written only once it is whole (#123) - 2026-08-14 Align the session codec's max-age with the 7-day absolute cap. The codec accepted cookies the session layer considered expired, so the cap was enforced in one place and not the other (#108)
- 2026-08-12 Warn when
TRUSTED_PROXIESis empty in production, where the safe default silently discards forwarded headers and every client rate-limits as the proxy's address (#149) - 2026-08-12 Bound the receiver rate limit per client IP across the
whole
/webhook/*route. The existing limiter keyed on the request path and/webhook/{uuid}matches any single segment, so a client that invented a fresh path per request minted a fresh bucket per request: the limit on the only unauthenticated endpoint bounded nothing in aggregate, and every request still cost an entrypoint lookup before it 404ed. An outer limiter keyed on the client address alone now bounds that, chained in front of the unchanged per-entrypoint limiter (#139) - 2026-08-12 Correct release-blocking documentation inaccuracies: the
README promised manual redelivery in the present tense in three
places when nothing implements it (the same false claim also sat in
the doc comment that was its source text), the env table omitted
RETENTION_SWEEP_INTERVAL, andTODO.mditself omitted five landed units (#141) - 2026-08-12 Make the CI gate execute the checks it reports on. The
workflow now writes a build-context fingerprint before calling
script/cibuild, so a code commit invalidates theCOPYlayer of the lint and builder stages while a docs-only commit still replays from cache; a superseding run also rewrites thefailurestatus Gitea leaves on commits it cancelled and never tested. Verified by pushing a deliberately broken test and watching CI go red (#119) - 2026-08-12 Require a positive
RETENTION_SWEEP_INTERVAL: a non-positive value reachedtime.NewTickerin both the retention reaper and the archive sweeper, panicking two goroutines with no recover after startup had already reported success (#140) - 2026-08-12 Bound the
X-Forwarded-Forscan's allocation to the hop cap: the reverse walk cuts entries withstrings.LastIndexByteinstead of joining and splitting, so a 1 MB header allocates 16 bytes rather than 1.6 MB per request on the unauthenticated receiver. Semantics proven unchanged by differential testing against the previous implementation (#133) - 2026-08-12 Cap the
X-Forwarded-Forhop walk at 64 entries, so an attacker-supplied chain cannot burn unbounded CPU in the rate-limit key function; running off the end falls back to the peer address (#124) - 2026-08-12 Gate forwarded-header trust behind a
TRUSTED_PROXIESCIDR list: all three rate limiters key on the connection's own address unless the direct peer is a configured proxy, in which caseX-Forwarded-Foris walked right to left for the first non-proxy hop. Default trusts nothing, and a set-but-unparseable value aborts startup. Before this, any client could mint a fresh bucket or drain another's by rotating a spoofed header (#88) - 2026-08-11 Web UI cleanup: nav terminology unified on Webhooks, the Profile settings placeholder removed, a progressive-enhancement copy button for the entrypoint URL, and retention form copy that states the actual policy (deletion by the reaper, 0 retains forever) (#57)
- 2026-08-11 Mask the webhook credential in delivery errors and logs:
Go embeds the request URL in
*url.Error, so every transport failure persisted the full Slack webhook URL into the per-webhook event database viaDeliveryResult.Error, a field a future REST API would have served.maskURLErrordrops path, query and userinfo while preserving the wrapped cause, soerrors.Is/AsandTimeout()still work and DNS, TLS and timeout failures still read differently (#118) - 2026-08-11 Rate-limit the public webhook receiver endpoint
(
RECEIVER_RATE_LIMIT, default 120/min), keyed on client IP plus entrypoint path so one entrypoint cannot exhaust another's budget; over-limit requests get 429 withRetry-After. It was the one unauthenticated, internet-facing endpoint with no limit at all (#64) - 2026-08-11 Enforce the body size limit before CSRF parses the form:
MaxBodySizeis now first in all four form-parsing route groups, so an oversized request is rejected with 413 instead of being read in full by the CSRF middleware before any cap applied (#90) - 2026-08-11 Mask target config on the source detail page, which
rendered the stored blob verbatim and so exposed the Slack
incoming-webhook URL — a bearer credential that cannot be revoked
per-holder. Config reaches the template only as a
TargetViewof labelled fields, and header values are rendered as a count (#113) - 2026-08-11 Allow
retention_daysof 0 to mean retain forever, via a sentinel written inBeforeSaveso the GORM column default cannot win the race. Also bounds the reaper's cutoff arithmetic: day counts above 106751 overflowedtime.Durationand wrapped the cutoff into the future, where every row matched and the sweep deleted everything (#79) - 2026-08-09 Inactivity-based session timeout: sliding idle expiry
(
SESSION_IDLE_TIMEOUT, default24h) refreshed on authenticated requests, with the 7-day absolute cap kept as an independent backstop that activity never extends (#66) - 2026-08-09 Restart recovery and the 60s retry sweep terminally fail an
orphaned
retryingdelivery whose target type no longer supports retries, recording aDeliveryResultwith the reason instead of leaving the delivery stuck forever (#82) - 2026-08-09 Root the delivery engine's worker pool and the retention
reaper's sweep loop at
context.Background()rather than the fxOnStarthook context (#97), which carries fx's 15s start timeout and killed both roughly fifteen seconds after boot: the proxy silently stopped delivering webhooks entirely, and the reaper never ran a single sweep under its default one-hour interval - 2026-08-09 Archive writer lifecycle (#89): deleting a webhook (or its
last
databasetarget) evicts the cached archive writer and closes its handle while deliberately leavingarchive-{webhookID}.dbon disk, and a newArchiveSweeperprunes idle archives on the existingRETENTION_SWEEP_INTERVALwithout ever creating an archive file - 2026-08-09 Configuration parsing fails loudly on set-but-unparseable
environment values:
envIntremoved in favour ofenvPositiveIntplus aPORTrange check,envBoolnow parses withstrconv.ParseBool, and defaults apply only to unset variables (#80) - 2026-08-07 Automatic event retention cleanup based on
retention_days, deleting expired events, deliveries, and delivery results from each per-webhook event database (#63) - 2026-08-07 Update golangci-lint to v2.12.2 (Docker image digest in
Dockerfile, release-archive sha256 pins inscript/bootstrap), adopt the canonical.golangci.yml(v2linters.settingslayout solll/funlen/cyclop/duplthresholds actually apply), and fix all newly surfaced lint findings - 2026-07-07 Adopted scripts-to-rule-them-all:
script/entrypoints, Makefile shims, README Entrypoints section - 2026-03-25 pin golangci-lint Docker image for linting (#55)
- 2026-03-18 CSRF middleware detects TLS per-request, fixing login over plain HTTP and behind reverse proxies (#54)
- 2026-03-17 root path redirects based on auth state (#52)
- 2026-03-17 CSRF protection, SSRF prevention for HTTP delivery targets with DNS rebinding defense, and per-IP login rate limiting (#42)
- 2026-03-17 Slack target type for incoming webhook notifications (#47)
- 2026-03-17 Dockerfile absolute paths and static linking (#49); absolute dev DATA_DIR default and clarified env docs (#46)
- 2026-03-05 security headers middleware, session regeneration on login, request body size limits (#41)
- 2026-03-04 tests for delivery, middleware, and session packages (#32); removed globals.Buildarch (#31)
- 2026-03-04 1.0 MVP merge: Webhook/Entrypoint/Target rename, core delivery engine with bounded worker pool and circuit breaker, parallel fan-out, per-webhook event databases, management UI (#16)
- 2026-03-01 repo brought to REPO_POLICIES standards; TODO.md folded into README (#6)
Future Steps
- Delivery status and retry management UI. Replay of a terminally failed delivery and per-attempt detail already landed (#203, #202)
- Per-webhook rate limiting in the receiver handler (per-webhook config plus handler enforcement; global limits must not apply to receiver endpoints)
- API key authentication for programmatic access (APIKey model exists; Bearer token middleware does not)
- REST API v1
- CRUD for webhooks, entrypoints, targets
- event viewing and filtering endpoints
- event redelivery endpoint
- OpenAPI specification
- Analytics dashboard: success rates, response times, volume
- A remember-me option at login
- Password reset flow for a forgotten password over the web. The
authenticated password change flow already landed, and a lost
password is recoverable from the console with
webhooker resetpw(#208) - Later, nice to have
- email delivery target type
- SNS and S3 delivery targets
- data transformations (e.g. webhook to Slack message formatting)
- JSONL file delivery with periodic S3 upload
- webhook event search and filtering
- multi-user with role-based access control