Address Reviews
Testing & Qualityaddress-reviewsAddress Reviews is an Agent Skill for testing & quality published in the FastLED/FastLED repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.
Fetch CodeRabbit review comments on the current PR, classify each finding, apply fixes for valid issues, and reply to each thread. Use before attempting gh pr merge when CodeRabbit has posted a review. Blocks merge on unresolved threads. Routes security-critical findings to a human.
How to use this skill
Bring this guide into your coding agent with a prompt tailored to the tool you use.
- Open your project in Codex.
- Copy and paste the prompt below.
- Review files and risks before approving any action.
Inspect the Agent Skill at https://github.com/FastLED/FastLED/blob/HEAD/.claude/skills/address-reviews/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/address-reviews, but wait for my approval before writing files or running scripts.
Copying this prompt does not install or run the skill. Review third-party files before use.
Address CodeRabbit review comments on a PR end-to-end: fetch, classify, fix, reply, commit, push. See ci/tools/coderabbit_addressor.py for the helper that drives the workflow.
Arguments: $ARGUMENTS
Workflow
-
Fetch comments. Run:
uv run ci/tools/coderabbit_addressor.py $ARGUMENTS --planThis prints a JSON plan listing every unresolved CodeRabbit comment, classified into one of four buckets:
valid-fix— actionable code changestyle— preference/nit; reply-onlyfalse-positive— disagree; reply with rationalesecurity-flag— flag a human; do not auto-fix
-
Review the plan. For each
valid-fix, read the file/line being flagged and decide the concrete change. Forfalse-positive, draft the reply. Do not touch anything insecurity-flag— leave it for a human reviewer. -
Apply fixes one comment at a time. Edit the referenced file, re-read to confirm the change is correct, and stage it.
-
Reply to each thread via:
uv run ci/tools/coderabbit_addressor.py $ARGUMENTS --reply <comment-id> "<message>"Keep replies terse: either "Fixed in " or a one-sentence rationale for disagreement.
-
Commit and push using a conventional-commit-format message that lists each addressed comment:
fix: address CodeRabbit review feedback - <comment-1-summary> - <comment-2-summary> -
Re-run the plan. If
valid-fixcount hits zero and nosecurity-flagremains, the PR is ready to merge. Otherwise iterate — hard limit: 3 passes.
Safety rails
- Max 3 iterations per PR. If CodeRabbit keeps asking for changes after 3 rounds, stop and flag the human reviewer — something structural is off.
- Never auto-fix
security-flagcomments. The classifier routes any comment mentioningsecurity,CVE,auth,credential,secret,RCE,injection,CRITICAL,data loss,UAF,use-after-free,buffer overflow, orout-of-boundsinto this bucket. - The pre-merge hook (
ci/hooks/check_pr_merge_reviews.py) blocksgh pr mergeuntil all CodeRabbit threads are resolved, so you cannot accidentally ship with pending feedback.
When to invoke
- Before any
gh pr mergeon a PR where CodeRabbit has posted review comments. - After pushing a new commit that might have triggered a re-review — the hook will remind you.
When NOT to invoke
- On draft PRs where CodeRabbit has not yet reviewed.
- When every open comment is already in a
security-flagstate — defer to a human.