tomsej/pi-ext/pr-review-comments
Claude Code and agent skill by tomsej/pi-ext
- community
Checking that the SKILL.md file is still in the repository.
Review, triage, and resolve PR review comments. Use when user asks to check PR comments, address review feedback, fix issues raised in review, reply to reviewers, or resolve review threads.
Install
npx vskill@latest install tomsej/pi-ext/pr-review-commentsSKILL.md
PR Review Comments
Triage and address review comments on a pull request: read each comment, decide if valid, fix code, verify tests pass, reply, and resolve threads.
Prerequisites
ghCLI authenticated- Working directory inside the repo (or use
--repo owner/repo)
Workflow
1. Fetch review comments
Get all review comments with context:
Also check top-level PR comments (non-inline):
2. Triage each comment
For each comment, decide:
- Valid & actionable → fix the code
- Valid but won't fix → reply with rationale
- Invalid / misunderstanding → reply with explanation
- Question → answer it
Read the relevant file context before deciding. Don't blindly accept or reject.
3. Fix code
Apply fixes for valid comments. Group related fixes into a single commit. Use conventional commit style:
Stage, commit, and push.
4. Validate tests pass
After fixing code, always run tests before pushing to make sure the fixes don't break anything:
- Run the project's test command (check project docs /
CLAUDE.md/Makefile/misetasks for the right command). - If tests fail, fix the issue before pushing. A review fix that breaks CI is worse than the original comment.
- If the fix changes test expectations (e.g., updated output values), update the tests too.
- Pay special attention to mock/stub setups — removing or changing mocks can cause unrelated tests to fail when test infrastructure intercepts more than expected (e.g.,
mock_providerintercepting locally-evaluated data sources).
Only push once all tests are green.
5. Reply to comments
Reply to each review comment thread using the in_reply_to field:
Important: The in_reply_to parameter must be a number (-F), not a string (-f). Use -F for numeric values.
Keep replies concise: state what was done or why not.
6. Resolve threads
Fetch unresolved thread node IDs via GraphQL:
Resolve each thread after replying:
Loop over all thread IDs to resolve in batch:
Rules
- Read before deciding. Always read the file/context around the commented line before accepting or rejecting.
- Don't blindly accept. Evaluate each comment on merit. Some automated/AI review comments are wrong.
- Don't blindly reject. If a comment points out a real issue, fix it even if the wording is off.
- Test before pushing. Run the full test suite (or at minimum the affected module's tests) after every fix. Never push code you haven't validated.
- Codify repeatable findings. If a fixed comment is mechanically enforceable or clearly recurring, add a guard in the same commit via the
review-guardsskill (ast-grep rule with test; qualitative → a line inREVIEW_GUIDELINES.md). - One commit for all fixes. Don't create separate commits per comment unless changes are unrelated.
- Reply to every comment. Even if resolved by a code change, confirm what was done.
- Resolve only after replying. Don't resolve threads without a reply — reviewers need to see acknowledgment.
- Push before replying. The fix commit should be visible in the PR when the reviewer reads the reply.
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/pr-review-comments/SKILL.md
Popularity
- Repository stars
- 78
- Repository forks
- 6
- 7-day trend score
- 22
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/tomsej/pi-ext/pr-review-comments)