Repository navigation
dev-lead: resolve review threads it has addressed — unresolved 'Applied in…' threads silently block merge via required_review_thread_resolution #1547
Description
Problem
Every pr-quality ruleset sets required_review_thread_resolution: true, so a PR cannot merge while any review thread is unresolved. The dev-lead review-changes flow applies fixes for bot review comments (gemini-code-assist, codeant-ai, CodeRabbit) and replies in-thread with "Applied in …" / "Verified and confirmed …" — but never resolves the thread. The PR then sits fully green + approved yet unmergeable, with nothing surfacing why.
During the 2026-08-18 compliance-PR sweep, 20 such threads had to be hand-resolved across ContentTwin #371/#424/#425, broodly #432/#483, bmad-bgreat-suite #429, TalkTerm #426/#427, and .github #970. In every case the fix was already live at head and the agent had already replied that it was addressed — resolution was the only missing step.
Requested fix
After the dev-lead posts an "addressed" reply to a review thread (and the fix commit is pushed), it should call the GraphQL resolveReviewThread mutation on that thread. Alternatively (or additionally) the pr-review cascade could resolve threads whose last reply is an agent addressed-marker when it verifies the fix at head.
Evidence
- 2026-08-18: resolving these threads immediately unblocked TalkTerm Evaluate Goal based criteria for dev-lead #427 (auto-merge fired within seconds) and was a prerequisite for 9 other merges.
🤖 Generated with Claude Code
Activity
don-petry commentedon Aug 20, 2026
Dev-Lead Implementation Plan
Issue: #1547 — dev-lead: resolve review threads it has addressed — unresolved 'Applied in…' threads silently block merge via required_review_thread_resolution
Scope
Add a deterministic, script-side safety net so that after the engine session, dev-lead resolves bot-originated review threads it has already addressed in-thread (the fix is at head, the agent replied but forgot to resolve) — the exact class of thread that sat unresolved and silently blocked merge under
required_review_thread_resolution: trueduring the 2026-08-18 sweep. The existingresolve_*_outdated_threadsnets only coverisOutdatedthreads; an addressed-at-head thread that GitHub has not marked outdated stays open. Resolution keys on a stable addressed-marker (<!-- dev-lead:addressed -->) so a skipped thread (no marker) is never resolved, and only bot-originated threads are touched — never a marker-less maintainer finding (respects the #1415 resolve-guard).Implementation Checklist
scripts/lib/maintainer-review-thread-gate.sh— add_DEV_LEAD_ADDRESSED_MARKERconstant +review_reply_is_addressed_marker <body>classifier (0 = carries the addressed-marker)scripts/dev-lead-fix-reviews.sh— addresolve_addressed_bot_threads <intent>: paginate review threads, resolve every unresolved bot-originated thread whose last comment carries the addressed-marker; dry-run + PR_NUMBER guards mirroringresolve_bot_outdated_threadsresolve_addressed_bot_threadsinto thefix-reviews,fix-bot-comment, andreview-changesintents alongside the existing resolve calls (applied + no-changes paths; skipped on the dev-lead fix-bot reverts its own fix on neutral Copilot 'overview' comments → net no-op PRs that falsely close compliance issues #1340 net-zero path)prompts/dev-lead/fix-reviews.md,fix-bot-comment.md,review-changes.md— instruct the agent to append<!-- dev-lead:addressed -->to every addressed replyTests to Write
tests/dev-lead/unit/test_maintainer_review_thread_gate.bats—review_reply_is_addressed_marker: marker present → 0; marker-less / skip reply / empty → 1tests/dev-lead/unit/test_fix_reviews.bats—resolve_addressed_bot_threads: dry-run announces + skips API; resolves an addressed bot thread (non-outdated) at head; does NOT resolve a marker-less bot thread; does NOT resolve a maintainer-originated threadtests/dev-lead/integration/test_prompt_coverage.shstays green (marker literal adds no undeclared${VAR})Test Command
bats tests/dev-lead/unit/test_fix_reviews.bats tests/dev-lead/unit/test_maintainer_review_thread_gate.bats+bash tests/dev-lead/integration/test_prompt_coverage.sh+bash .dev-lead/scripts/dev-lead-lint.shdon-petry commentedon Aug 20, 2026
Dev-Lead: Progress (in progress — not yet durable)
Plan Execution (intended)
required_review_thread_resolution: true(the 2026-08-18 sweep hand-resolved 20).Files touched (this workspace, not yet pushed)
scripts/lib/maintainer-review-thread-gate.sh: added the<!-- dev-lead:addressed -->addressed-marker constant +review_reply_is_addressed_markerclassifier (single source of truth for the marker).scripts/dev-lead-fix-reviews.sh: addedresolve_addressed_bot_threads(paginated; resolves an unresolved bot-originated thread whose last reply carries the addressed-marker; dry-run +PR_NUMBERguards) and wired it into thefix-reviews,fix-bot-comment, andreview-changesintents alongside the existing outdated-thread nets (skipped on the dev-lead fix-bot reverts its own fix on neutral Copilot 'overview' comments → net no-op PRs that falsely close compliance issues #1340 net-zero path).prompts/dev-lead/{fix-reviews,fix-bot-comment,review-changes}.md: instruct the agent to end each addressed reply with<!-- dev-lead:addressed -->(only on genuine addressed replies, never on skips).tests/dev-lead/unit/test_maintainer_review_thread_gate.bats(classifier) andtests/dev-lead/unit/test_fix_reviews.bats(dry-run announce; resolves a non-outdated addressed bot thread; leaves a marker-less bot thread open; leaves a maintainer-originated thread open).Notes
resolve_bot_outdated_threads, which only coversisOutdatedthreads; an addressed-at-head finding is frequently not marked outdated by GitHub, which is exactly the gap that stuck the sweep PRs.scripts/dev-lead-lint.sh, prompt-coverage integration test, and the dev-lead unit+integration bats suites all ran clean in this workspace.don-petry commentedon Aug 20, 2026
Dev-Lead: Implementation Complete — PR #1555
The implementation for issue #1547 is durable: commits are pushed and a pull request is open. This record is backed by verifiable artifacts you can confirm in one click.
d9bfe4c28625260346d207fa9d9b0ee4e8e58098Review 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.
feat: implement issue #1547 — dev-lead: resolve review threads it has…