Milestoned 1.0.0. Docs-only, but it actively misleads and it will cause churn against work landing right now.
REPO_POLICIES.md declares in its front matter that it is synced from https://git.eeqj.de/sneak/prompts. The local copy is stale — last_modified: 2026-07-06 against upstream's 2026-08-07 — and has drifted in three substantive places, all of them upstream being newer:
The test-duration policy. Local says "make test must complete in under 20 seconds. Add a 30-second timeout in the Makefile." Upstream (commit a868689, prompts PR #42, "Raise org-wide make test cap to 60s, backstop timeout to 90s") says the hard cap is 60 s, the target is 20 s, a suite between 20 and 60 s is still green but its overage must be filed as an improvement bug, and the Makefile backstop is 90 s — deliberately above the hard cap so it catches a hung test rather than a merely slow one.
The Makefile example in the same section, -timeout 30s in both the plain and verbose-rerun invocations, against upstream's -timeout 90s.
The golangci-lint section, which upstream extends with the canonical version (v2.12.2, released 2026-05-06) and its commit-pinned go install line.
Nothing else differs; the whole delta is 30 diff lines.
Why this is worth fixing before the tag rather than after
It has already caused a concrete error. Working #194 I read this file, concluded the org policy mandated 30 s, and was about to open a speculative PR against sneak/prompts proposing a change that had already been made — nearly spending someone's review attention on a question that was closed a fortnight ago.
It will also cause churn in the other direction: #195 raises this repo's test timeout, and a future compliance pass reading this stale file would revert it to 30 s and reintroduce the defect 194 exists to fix. That PR carries a note in script/test guarding against exactly that, which stops being necessary once this file is correct.
Definition of done
REPO_POLICIES.md matches upstream prompts/REPO_POLICIES.md — take upstream wholesale rather than merging hunk by hunk, since this file is a synced copy and not a place for local edits.
Check whether anything in this repo depended on the stale text before assuming a copy is safe: the timeout figure in particular is referenced by script/test. Report what you found either way.
If upstream has moved again by the time you do this, sync to whatever upstream is then, and say which commit.
Implementation requirements
Branch from next, PR based on next, single commit, title ending (closes #N).
Docs-only: if the change touches nothing but REPO_POLICIES.md, no adversarial review is needed. If you find yourself changing code or script/, stop — that is a different unit.
Run make fmt and include the result; never commit unformatted markdown. Gate on make check.
Milestoned **1.0.0**. Docs-only, but it actively misleads and it will cause churn against work landing right now.
`REPO_POLICIES.md` declares in its front matter that it is synced from https://git.eeqj.de/sneak/prompts. The local copy is stale — `last_modified: 2026-07-06` against upstream's `2026-08-07` — and has drifted in three substantive places, all of them upstream being newer:
1. **The test-duration policy.** Local says "`make test` must complete in under 20 seconds. Add a 30-second timeout in the Makefile." Upstream (commit `a868689`, prompts PR #42, "Raise org-wide `make test` cap to 60s, backstop timeout to 90s") says the hard cap is 60 s, the target is 20 s, a suite between 20 and 60 s is still green but its overage must be filed as an improvement bug, and the Makefile backstop is **90 s** — deliberately above the hard cap so it catches a hung test rather than a merely slow one.
2. **The Makefile example** in the same section, `-timeout 30s` in both the plain and verbose-rerun invocations, against upstream's `-timeout 90s`.
3. **The golangci-lint section**, which upstream extends with the canonical version (v2.12.2, released 2026-05-06) and its commit-pinned `go install` line.
Nothing else differs; the whole delta is 30 diff lines.
## Why this is worth fixing before the tag rather than after
It has already caused a concrete error. Working https://git.eeqj.de/sneak/webhooker/issues/194 I read this file, concluded the org policy mandated 30 s, and was about to open a speculative PR against `sneak/prompts` proposing a change that had already been made — nearly spending someone's review attention on a question that was closed a fortnight ago.
It will also cause churn in the other direction: https://git.eeqj.de/sneak/webhooker/pulls/195 raises this repo's test timeout, and a future compliance pass reading this stale file would revert it to 30 s and reintroduce the defect 194 exists to fix. That PR carries a note in `script/test` guarding against exactly that, which stops being necessary once this file is correct.
## Definition of done
- `REPO_POLICIES.md` matches upstream `prompts/REPO_POLICIES.md` — take upstream wholesale rather than merging hunk by hunk, since this file is a synced copy and not a place for local edits.
- **Check whether anything in this repo depended on the stale text** before assuming a copy is safe: the timeout figure in particular is referenced by `script/test`. Report what you found either way.
- If upstream has moved again by the time you do this, sync to whatever upstream is then, and say which commit.
## Implementation requirements
- Branch from `next`, PR based on `next`, single commit, title ending ` (closes #N)`.
- Docs-only: if the change touches nothing but `REPO_POLICIES.md`, no adversarial review is needed. If you find yourself changing code or `script/`, stop — that is a different unit.
- Do not modify `TODO.md` (see https://git.eeqj.de/sneak/webhooker/issues/112).
- Run `make fmt` and include the result; never commit unformatted markdown. Gate on `make check`.
clawbot
self-assigned this 2026-08-18 09:31:16 +02:00
clawbot
added this to the 1.0.0 milestone 2026-08-18 09:31:16 +02:00
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Milestoned 1.0.0. Docs-only, but it actively misleads and it will cause churn against work landing right now.
REPO_POLICIES.mddeclares in its front matter that it is synced from https://git.eeqj.de/sneak/prompts. The local copy is stale —last_modified: 2026-07-06against upstream's2026-08-07— and has drifted in three substantive places, all of them upstream being newer:make testmust complete in under 20 seconds. Add a 30-second timeout in the Makefile." Upstream (commita868689, prompts PR #42, "Raise org-widemake testcap to 60s, backstop timeout to 90s") says the hard cap is 60 s, the target is 20 s, a suite between 20 and 60 s is still green but its overage must be filed as an improvement bug, and the Makefile backstop is 90 s — deliberately above the hard cap so it catches a hung test rather than a merely slow one.-timeout 30sin both the plain and verbose-rerun invocations, against upstream's-timeout 90s.go installline.Nothing else differs; the whole delta is 30 diff lines.
Why this is worth fixing before the tag rather than after
It has already caused a concrete error. Working #194 I read this file, concluded the org policy mandated 30 s, and was about to open a speculative PR against
sneak/promptsproposing a change that had already been made — nearly spending someone's review attention on a question that was closed a fortnight ago.It will also cause churn in the other direction: #195 raises this repo's test timeout, and a future compliance pass reading this stale file would revert it to 30 s and reintroduce the defect 194 exists to fix. That PR carries a note in
script/testguarding against exactly that, which stops being necessary once this file is correct.Definition of done
REPO_POLICIES.mdmatches upstreamprompts/REPO_POLICIES.md— take upstream wholesale rather than merging hunk by hunk, since this file is a synced copy and not a place for local edits.script/test. Report what you found either way.Implementation requirements
next, PR based onnext, single commit, title ending(closes #N).REPO_POLICIES.md, no adversarial review is needed. If you find yourself changing code orscript/, stop — that is a different unit.TODO.md(see #112).make fmtand include the result; never commit unformatted markdown. Gate onmake check.clawbot referenced this issue2026-08-18 10:50:50 +02:00