Skip to content

docs: read the reviewer's status notice when a review loop is stuck#88

Merged
thejustinwalsh merged 1 commit into
mainfrom
chore/review-bot-notices
May 24, 2026
Merged

docs: read the reviewer's status notice when a review loop is stuck#88
thejustinwalsh merged 1 commit into
mainfrom
chore/review-bot-notices

Conversation

@thejustinwalsh

@thejustinwalsh thejustinwalsh commented May 24, 2026

Copy link
Copy Markdown
Owner

Summary

Refines the "bump when stuck" guidance (skill review-loop step 4 + Phase 5 spec resume step). The naive rule — "comment @coderabbitai review" — is wrong for the most common stall: CodeRabbit auto-pauses after an influx of commits (which the batch-and-push review loop itself produces), and a one-shot @coderabbitai review does not un-pause it. The bot posts a status notice naming the exact command.

New rule: when a review is stuck, read the reviewer's status notice and run the command it names@coderabbitai resume (re-enable auto-review), @coderabbitai review (single pass), @coderabbitai resolve (close addressed threads). Don't guess.

Caught live: #86's review was paused by our own loop's commit churn; @coderabbitai resume (not review) is what brought it back green.

Skill change propagated across all four copies (canonical / mirror / .claude / .codex); drift gate passes.

Summary by CodeRabbit

  • Documentation
    • Enhanced guidance for handling stalled bot reviews with improved instructions on checking reviewer status notices and using targeted commands (@coderabbitai resume, @coderabbitai review, @coderabbitai resolve) to unblock workflows. Added clarity on distinguishing between paused and active review states.

Review Change Stack

@coderabbitai

coderabbitai Bot commented May 24, 2026

Copy link
Copy Markdown

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info
⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Pro Plus

Run ID: 9e17ade2-a01f-47d8-95a2-7508dc2f3b62

📥 Commits

Reviewing files that changed from the base of the PR and between aaa9a85 and 6c2809d.

📒 Files selected for processing (5)
  • .claude/skills/implementing-github-issues/SKILL.md
  • .codex/skills/implementing-github-issues/SKILL.md
  • packages/cli/src/bootstrap-assets/skills/implementing-github-issues/SKILL.md
  • packages/skills/implementing-github-issues/SKILL.md
  • planning/middle-management-build-spec.md

📝 Walkthrough

Walkthrough

This PR synchronizes documentation across skill guides and build specifications to clarify handling when bot review becomes stalled. The workflow now directs users to check the reviewer's latest status notice and execute the appropriate CodeRabbit command (resume, review, or resolve) rather than blindly retrying with a single command.

Changes

Review Loop Bot-State Handling

Layer / File(s) Summary
Stuck review state handling across documentation
.claude/skills/implementing-github-issues/SKILL.md, .codex/skills/implementing-github-issues/SKILL.md, packages/cli/src/bootstrap-assets/skills/implementing-github-issues/SKILL.md, packages/skills/implementing-github-issues/SKILL.md, planning/middle-management-build-spec.md
All documentation files updated with consistent guidance: when review status doesn't advance after pushing changes, read the reviewer's posted status notice and select the correct @coderabbitai command. resume resumes paused automation; review runs a fresh single pass; resolve handles addressed threads. Build spec adds agent respawning logic with full review-pass threads and round-cap enforcement.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~12 minutes

Possibly related issues

  • thejustinwalsh/middle#32: Directly implements the Phase 5 review-resume logic with explicit @coderabbitai resume/review/resolve commands and reviewer-status handling described in this PR's spec updates.
  • thejustinwalsh/middle#36: Requires agent resume for review-changes including the internal review loop and CodeRabbit command logic now clarified in this PR's skill documentation.

Possibly related PRs

  • thejustinwalsh/middle#77: Updates the same Phase 5 review-changes resume logic in the build spec to refine agent dispatch when review status is stuck.
  • thejustinwalsh/middle#87: Updates the implementing-github-issues skill's review-loop guidance for handling stuck CHANGES_REQUESTED and CodeRabbit command workflows.

Suggested labels

ready-for-review

🚥 Pre-merge checks | ✅ 5
✅ Passed checks (5 passed)
Check name Status Explanation
Description Check ✅ Passed Check skipped - CodeRabbit’s high-level summary is enabled.
Title check ✅ Passed The title directly and specifically describes the main change: documenting that developers should read the reviewer's status notice when a review loop stalls, which is the core improvement across all four skill/spec files.
Docstring Coverage ✅ Passed No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Linked Issues check ✅ Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check ✅ Passed Check skipped because no linked issues were found for this pull request.

✏️ Tip: You can configure your own custom pre-merge checks in the settings.


Comment @coderabbitai help to get the list of available commands and usage tips.

@thejustinwalsh thejustinwalsh merged commit 91453e2 into main May 24, 2026
1 check passed
thejustinwalsh added a commit that referenced this pull request May 24, 2026
…hase 4 gates + #80

Phase 5 (#32) was built against a pre-Phase-4 base; main since landed the
skill-enforcement gates (#86), the positive-done-signal nudge loop (#80), the
plan-comment guard, and the review-handling skill codification (#87/#88/#89).
Integration (Phase 5 as the base, main's additions layered on):

- driveOnce resolves a bare-stop through the #80 nudge loop (session stays alive
  during nudges); a ready, non-draft Epic PR resolves to `done`, which Phase 5
  then parks on review-resolved. nudge-exhausted parks in waiting-human.
- The plan-comment guard runs in driveOnce: an unposted `done` is demoted to
  `failed` so it never enters the review park.
- DriveResult carries a DriveOutcome (StopClassification | nudge-exhausted);
  finalStateFor handles nudge-exhausted; the drive-step timeout widened for nudges.
- Skill (4 copies) resolved to main's codified version; the #80 nudge +
  plan-comment tests rewritten park-aware.

Verified: typecheck clean, full suite 289 pass / 0 fail, skills mirror in sync.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant