← ClaudeAtlas

resolving-pr-commentslisted

Resolves automated PR review bot findings end-to-end: fetch the bot comment, fix every Must-fix, best-effort Should-fix, fold in Nits already in-region, run gates, reconcile docs, commit, push, reply. Use for "fix what the bot flagged", "claude-pr-review posted FAIL", "address PR feedback". Skip manual reviews/generic CI failures. GitHub stance: hub-only, GitHub MCP (ADR-0030).
monte3l/m3l-automation · ★ 0 · AI & Automation · score 56
Install: claude install-skill monte3l/m3l-automation
# resolving-pr-comments Automates the loop of reading the `claude-pr-review.yml` bot's findings, fixing each one, verifying quality gates, and closing the loop with a reply comment — so you spend zero time on mechanical review-driven edits. ## Boundary rules - Never push with `--force`. - If any quality gate fails after your fixes, **stop and report the failure** — do not commit or push. - If a **Must-fix** finding requires a structural change you cannot make as a targeted line fix (e.g., redesigning an entire type hierarchy, splitting a test suite), describe what is needed and ask the user to handle it before continuing — Must-fix blocks merge, so it cannot be silently skipped. - If a **Should-fix** finding requires the same kind of structural change, skip it instead of asking — it does not block merge, so stopping the whole run over an optional item is worse than leaving it for a human. Note it as not addressed in the Step 3 preview, the Step 9 commit body, and the Step 10 follow-up comment — skipping the fix never means skipping the acknowledgment: an `Acknowledged-Should-Fix:` footer still lands on a commit in this pass (Step 9), because `should-fix-ack` (`docs/adr/0097`) requires one whenever any Should-fix was posted, fixed or not. - For a finding you believe is wrong, this skill does not dispute it on your behalf — see `docs/contributing/branch-protection.md`'s override procedure (investigate, reply to the bot's comment thread with evidence,