Repository navigation
Description
Summary
dev-lead posts its acknowledgement of a non-actionable bot notice as a plain, unmarked PR issue comment. The maintainer-comment gate (#1290/#1813) classifies comments by marker, not by author — and dev-lead commits and comments as don-petry, a human login. So each acknowledgement is indistinguishable from a maintainer finding and becomes a new undispositioned comment that blocks the very approval it was posted to unblock.
The gate is not wrong and dev-lead is not wrong. The two compose into a loop that can only ever add blockers.
Evidence, 2026-09-24
On PR #1887, the unminimized, unmarked comments authored by don-petry were all dev-lead's own dispositions:
Acknowledged — this is a Qodo trial-ended/billing notice, not a code finding. No action needed
Acknowledged — this is a Codex usage-limit notice for code reviews, not a code finding. No acti
Acknowledged — this is a CodeRabbit rate-limit notice ("Review limit reached"), not an actionab
Acknowledged — SonarQube Cloud reports the Quality Gate **passed**. …
Acknowledged — SonarQube Cloud reports the Quality Gate **passed** for …
Five acknowledgements, five fresh gate blockers. The same pattern is on #1873 (three) and others. Every one of dev-lead's marker-prefixed comments (<!-- dev-lead-fix-reviews … -->, <!-- dev-lead-hold-notice … -->, <!-- dev-lead rate-limit-ack -->) is correctly excluded — it is only the unmarked prose acknowledgements that leak through.
Note the irony in the list above: three of the five acknowledge third-party reviewer rate-limit notices, which are themselves a consequence of the advisory bots being out of quota. The fleet generates noise, acknowledges the noise, and the acknowledgement blocks the merge.
Why this matters more than it looks
Approval on these PRs is already scarce — see #1894, where ten of eleven open PRs had never received an approving review. This loop means the longer dev-lead works a PR, the more gate blockers accumulate on it, and none of them represent a real finding. Before #1910/#1911 shipped scripts/maintainer-resolve-comment.sh there was no way to clear them without dev-lead, so they were permanent.
Acceptance Criteria
- dev-lead's own acknowledgement/disposition comments carry an agent marker (the same convention its other comments already use), so the maintainer-comment gate excludes them.
- Existing unmarked acknowledgements already on open PRs are cleared, or the gate is taught to recognise them, so the current queue is not left carrying them.
scripts/maintainer-resolve-comment.shis the sanctioned path for an agent-authored comment under the same identity. - A test asserts that a dev-lead acknowledgement does not increment the gate's undispositioned count. This is the regression that matters — the marker convention is easy to forget on a new comment type.
- Consider whether an acknowledgement needs to be a PR comment at all. Five comments saying "a third-party bot is rate-limited, no action needed" is noise on the PR regardless of the gate; a single collapsed marker comment, or a step-summary line, may serve better.
Notes for the implementer
- Do not fix this by making the gate author-aware. dev-lead deliberately acts as
don-petry, so "exclude comments by this login" would also exclude every genuine maintainer finding — the exact defect PR comments from maintainers are mechanically inert — they neither block merge nor reach dev-lead #1290 exists to prevent. The marker is the correct discriminator; the acknowledgements simply need to use it. - Check every dev-lead comment-emitting path for the same omission, not just the bot-notice acknowledgement. The marker-prefixed paths are already correct, so this is about finding the ones that bypassed the convention.
References
- PR comments from maintainers are mechanically inert — they neither block merge nor reach dev-lead #1290 / pr-review can't approve ANY open PR: maintainer-comment gate needs pushedDate (always null) and treats codeant/qodo/auto-rebase comments as maintainer findings — 18/18 PRs stranded #1813 — the maintainer-comment gate and its verified-disposition redesign
- 10 of 11 open PRs have never received an approving review: the advisory quorum is unmeetable, the disposition gate revokes what does issue, and nothing reports why #1894 — ten of eleven PRs never approved; this is one contributing mechanism
- #1894 AC #4: give maintainers a dev-lead-independent way to disposition a blocking PR comment #1910 / PR feat: implement issue #1910 — #1894 AC #4: give maintainers a dev-lead-independent way to disposition a blocking PR comment #1911 —
maintainer-resolve-comment.sh, the tool that makes existing ones clearable scripts/lib/maintainer-comment-gate.sh—_MAINTAINER_GATE_AGENT_MARKERS
Activity
github-actions commentedon Sep 24, 2026
Thanks for the detailed write-up and the concrete per-PR evidence — the loop is clearly described and the five acknowledgements on #1887 make a compelling reproduction case.
One question to help scope the fix: have you already identified the specific script paths or functions that emit these unmarked acknowledgement comments? Knowing which emit-paths are missing the marker (versus the ones that already use
<!-- dev-lead-... -->correctly) would let an implementer go straight to the right places rather than auditing the full scripts directory first.don-petry commentedon Sep 24, 2026
Cross-reference: #1918 is the same failure shape for bot comments (SonarCloud's re-posted "Quality Gate passed"), and is further along with evidence that #1913 was approved twice and dismissed both times. This issue covers dev-lead's own unmarked acknowledgements.
Both reduce to the same root: the gate's only satisfying signal is a
RESOLVEDminimize, and fewer actors can produce one than can post a blocking comment. Please solve them together — fixing either alone leaves the queue stranded. Verified on #1887: clearing all five dev-lead acknowledgements by hand still yieldedapprovals=0from a completed pr-review run, because onesonarqubecloudcomment remained.don-petry commentedon Sep 24, 2026
Dev-Lead Implementation Plan
Issue: #1919 — dev-lead's own unmarked acknowledgement comments block the approval they exist to unblock
Root cause
The maintainer-comment gate (
scripts/lib/maintainer-comment-gate.sh) classifies PR issue comments by marker, not author. Its regex already excludes any<!-- dev-lead … -->comment. dev-lead's marked paths (fix-reviews Phase 1b disposition, hold-notice, rate-limit-ack) are correctly excluded. Two comment-emitting paths bypass the convention and post unmarked PR issue comments that then count as undispositioned blockers:prompts/dev-lead/review-changes.md(§54–60) — the failing-check fix comment posted viagh pr comment, no marker.prompts/dev-lead/fix-bot-comment.md— freeform "Acknowledged — …" acks of non-actionable bot issue-comment notices (the five in the evidence), no marker and arguably shouldn't be comments at all.Scope
Teach every dev-lead PR-issue-comment path to stamp a
<!-- dev-lead … -->marker (so the existing gate excludes it), stop posting pure-noise bot-notice acks as comments, and add the gate regression test.Implementation Checklist
prompts/dev-lead/review-changes.md— failing-check fix comment must end with<!-- dev-lead:check-fix -->(AC1)prompts/dev-lead/fix-bot-comment.md— for a non-actionable bot issue-comment notice, prefer a run/step-summary line over a PR comment; if a PR comment is posted it MUST carry a<!-- dev-lead:ack -->marker (AC1 + AC4).github/workflows/lint.yml— register the new prompt-convention bats file in the bats listTests to Write
tests/dev-lead/unit/test_maintainer_comment_gate.bats— a dev-lead-marked ack (<!-- dev-lead:ack -->) and a<!-- dev-lead:check-fix -->comment do not increment the gate's undispositioned count (→ 0); an UNMARKED "Acknowledged — …" prose comment still blocks (→ 1). This is AC3 — the regression that matters.tests/dev-lead/unit/test_agent_marker_convention.bats— assert every PR-issue-comment-posting prompt path instructs stamping a<!-- dev-lead … -->marker (guards against forgetting it on a new comment type)Not code changes
scripts/maintainer-resolve-comment.shalready resolves a comment authored under dev-lead's own identity (don-petry). New acks carry the marker, so no clearing is needed for them.Test Command
bats tests/dev-lead/unit/test_maintainer_comment_gate.bats tests/dev-lead/unit/test_agent_marker_convention.batsandbash .dev-lead/scripts/dev-lead-lint.shdon-petry commentedon Sep 24, 2026
Dev-Lead: Progress (in progress — not yet durable)
Root cause (confirmed)
The maintainer-comment gate discriminates PR issue comments by marker, not author. dev-lead's marked paths (fix-reviews Phase 1b disposition, hold-notice, rate-limit-ack) are already excluded. Two paths posted unmarked PR issue comments that then counted as undispositioned blockers:
prompts/dev-lead/review-changes.md— the failing-check fix note (gh pr comment), no marker.prompts/dev-lead/fix-bot-comment.md— freeform "Acknowledged — …" acks of non-actionable bot notices, no marker (the five in the evidence).Plan execution (intended)
review-changes.mdfailing-check comment with<!-- dev-lead:check-fix -->(the gate'sdev-leadmarker regex already excludes it).fix-bot-comment.mdto not post an unmarked acknowledgement of a non-actionable bot notice: prefer the run/step summary (AC4), and if a PR comment is genuinely warranted it must carry<!-- dev-lead:ack -->.:ack/:check-fix-marked comment does not increment the undispositioned count (→ 0), while the exact unmarked "Acknowledged — …" prose still blocks (→ 1) — the marker, not the prose, is what clears it (AC3).tests/dev-lead/unit/test_agent_marker_convention.bats(registered inlint.yml) asserting every PR-issue-comment prompt path instructs the marker.On AC2 (existing unmarked acks on open PRs)
No code change — operational.
scripts/maintainer-resolve-comment.shalready resolves a comment authored under dev-lead's own identity (don-petry); new acks carry the marker, so only the pre-existing ones need that one-time clear.Local check output (provisional — not the pushed run)
Files touched (this workspace, not yet pushed)
prompts/dev-lead/review-changes.md,prompts/dev-lead/fix-bot-comment.mdtests/dev-lead/unit/test_maintainer_comment_gate.bats,tests/dev-lead/unit/test_agent_marker_convention.bats.github/workflows/lint.ymlNotes
No change to the gate itself — the
<!-- dev-lead … -->exclusion the issue relies on already exists; the fix is making the emitting paths honour the convention (per the issue's "do not make the gate author-aware" guidance). Sibling #1918 (bot status comments) shipped separately in #1920.don-petry commentedon Sep 24, 2026
Dev-Lead: Implementation Complete — PR #1928
The implementation for issue #1919 is durable: commits are pushed and a pull request is open. This record is backed by verifiable artifacts you can confirm in one click.
7ac240344cf1fc0994362eae40f80ac236e7d86cmain: 5 file(s), +87/-1 linesReview happens on the PR. Acceptance criteria and test results are evidenced by the PR's diff and CI checks — not asserted here. If a later stage fails, this claim is retracted in place.
fix(dev-lead): target the triggering bot notice by node id; never put…
fix(dev-lead): case 1 needs the full clean-status pattern; one marker…
fix(dev-lead): accept padded node ids, fail loudly without one, bind …
fix(dev-lead): a SonarCloud pass-with-findings is never dispositioned…
feat: implement issue #1919 — dev-lead's own unmarked acknowledgement…