pr-reviewer
Skill326 starsChanged 8 months agoMITView on GitHub
Fetches PR review comments, analyzes requested changes, implements fixes, commits and pushes the resolution. Use after a reviewer has left comments on an open PR to close the feedback loop efficiently.
- Deletes or force-pushes
- Commits and pushes
22 other projects use a version of this.
Adopted by
Most recently taken up by
- idavidov13/agentic-playwrighttheir own version, first seen Nov 8, 2026
- getpapi/papitheir own version, first seen Nov 1, 2026
- kcayme/k-agentstheir own version, first seen Nov 1, 2026
- odykyi/gh-skillstheir own version, first seen Oct 25, 2026
- archiforge/agent-crafttheir own version, first seen Oct 20, 2026
- jpantsjoha/ai-native-developer-experiencetheir own version, first seen Oct 16, 2026
Other versions22 projects use a version of this.
Word for word · 1
- Fu-Jie/awesome-openwebui★ 326100% same
Edited · 0
None found.
Wrote their own version · 21
- pyuvm/pyuvm★ 574same name
- idavidov13/agentic-playwright★ 268same name
- proflead/codex-skills-library★ 153same name
- h0tp-ftw/ankimon★ 83same name
- glincker/claude-code-marketplace★ 37same name
- jpantsjoha/ai-native-developer-experience★ 13same name
- getpapi/papi★ 11same name
- tstapler/dotfiles★ 8same name
- gvre/skills★ 7same name
- k3ybladewielder/skills★ 7same name
11 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository16 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- .github/copilot-instructions.mdCopilot instructions
- community-announcerSkill
- doc-mirror-syncSkill
- gh-discussion-replierSkill
- gh-issue-replierSkill
- gh-issue-schedulerSkill
- i18n-validatorSkill
- playwright-openwebuiSkill
- plugin-scaffolderSkill
- pr-submitterSkill
- publish-no-version-bumpSkill
- release-finalizerSkill
- release-prepSkill
- source-code-analyzerSkill
- version-bumperSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
13 rules. 0 are common, 8 are distinctive, and 5 are very specific to this project.
0
Common
Most of the 21 edited versions we read keep these.
8
Distinctive
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Distinctive: line 44
If mismatched, checkout the correct branch first.
- Distinctive: line 65
Read the affected file at the commented line for context:
- Distinctive: line 70
After editing, verify the specific area looks correct
- Very specific: line 73
Check if tests exist: ls tests/test_*.py
- Very specific: line 74
If tests exist, run them: python -m pytest tests/ -v
- Very specific: line 77
If modifying README.md, check if docs/ mirror needs the same fix
- Distinctive: line 82
After all fixes are applied:
- Distinctive: line 94
Create a new commit (do NOT amend if the branch has already been pushed, to avoid force-push):
- Distinctive: line 125
Use git push (non-force) by default
- Distinctive: line 126
Only use git push --force-with-lease if:
- Very specific: line 129
Never use --force (without --lease)
- Very specific: line 149
If the reviewer had submitted a CHANGES_REQUESTED review, request a new review after fixes:
- Distinctive: line 164
Large refactors: If fix would require changing >50 lines, propose a separate follow-up issue instead
“Common” is measured: kept by at least 6 in 10 of the 21 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
PR Reviewer
Overview
This skill automates the response cycle for code review. When a reviewer leaves comments on a Pull Request, this skill fetches all pending feedback, categorizes issues by severity, implements fixes, and submits a follow-up commit with appropriate review response comments.
Prerequisites
- An open PR exists with pending review comments
- The local branch matches the PR's head branch
ghCLI is authenticated
Workflow
Step 1 — Fetch Review State
Retrieve all review comments and overall review status:
# Get overall review decisions
PAGER=cat GH_PAGER=cat gh pr view <PR-NUMBER> --json reviews,reviewDecision,headRefName \
--jq '{decision: .reviewDecision, reviews: [.reviews[] | {author: .author.login, state: .state, body: .body}]}'
# Get inline code comments (specific line comments)
PAGER=cat GH_PAGER=cat gh api repos/Fu-Jie/openwebui-extensions/pulls/<PR-NUMBER>/comments \
--jq '[.[] | {path: .path, line: .line, body: .body, author: .user.login, id: .id}]'
# Get general issue comments
PAGER=cat GH_PAGER=cat gh issue view <PR-NUMBER> --comments --json comments \
--jq '[.comments[] | {author: .author.login, body: .body}]'Confirm the current local branch matches the PR head:
git branch --show-currentIf mismatched, checkout the correct branch first.
Step 2 — Categorize Review Feedback
Group feedback into categories:
| Category | Examples | Action |
| Code Bug | Logic error, incorrect variable, broken condition | Fix code immediately |
| Style / Formatting | Indentation, naming convention, missing blank line | Fix code |
| Documentation | Missing i18n key, wrong version in README, typo | Fix docs |
| Design Question | Suggestion to restructure, alternative approach | Discuss with user before implementing |
| Nitpick / Optional | Minor style preferences reviewer marked as optional | Fix if quick; document if skipped |
| Blocking | Reviewer explicitly blocks merge | Must fix before proceeding |
Present the full categorized list to the user and confirm the resolution plan.
Step 3 — Implement Fixes
For each accepted fix:
- Read the affected file at the commented line for context:
sed -n '<line-5>,<line+10>p' <file-path> - Apply the fix using appropriate file edit tools
- After editing, verify the specific area looks correct
For code changes that might affect behavior:
- Check if tests exist:
ls tests/test_*.py - If tests exist, run them:
python -m pytest tests/ -v
For documentation fixes:
- If modifying README.md, check if
docs/mirror needs the same fix - Apply the same fix to both locations
Step 4 — Run Consistency Checks
After all fixes are applied:
# Version consistency (if any version files were touched)
python3 scripts/check_version_consistency.py
# Quick syntax check for Python files
python3 -m py_compile plugins/{type}/{name}/{name}.py && echo "✅ Syntax OK"Step 5 — Stage and Commit
Create a new commit (do NOT amend if the branch has already been pushed, to avoid force-push):
git add -A
git statusDraft a Conventional Commits message for the fixup:
Format: fix(scope): address review feedback
Body should list what was fixed, referencing reviewer concerns:
fix(github-copilot-sdk): address review feedback from @reviewer
- Fix X per review comment on line Y of file Z
- Update README to clarify auth requirement
- Correct edge case in _parse_mcp_servers logicgit commit -m "<fixup commit message>"Step 6 — Push the Fix Commit
git push origin $(git branch --show-current)Force-push policy:
- Use
git push(non-force) by default - Only use
git push --force-with-leaseif:- The user explicitly requests it, AND
- The only change is an amended commit squash (cosmetic, no logic change)
- Never use
--force(without--lease)
Step 7 — Respond to Reviewers
For each addressed review comment, post a reply:
# Reply to inline comment
gh api repos/Fu-Jie/openwebui-extensions/pulls/<PR-NUMBER>/comments/<COMMENT-ID>/replies \
-X POST -f body="Fixed in commit <SHORT-SHA>. <Brief explanation of what was changed.>"
# General comment to summarize all fixes
gh issue comment <PR-NUMBER> --body "All review feedback addressed in commit <SHORT-SHA>:
- Fixed: <item 1>
- Fixed: <item 2>
Ready for re-review. 🙏"Step 8 — Re-Request Review (Optional)
If the reviewer had submitted a CHANGES_REQUESTED review, request a new review after fixes:
PAGER=cat GH_PAGER=cat gh api repos/Fu-Jie/openwebui-extensions/pulls/<PR-NUMBER>/requested_reviewers \
-X POST -f reviewers[]='<reviewer-login>'Decision Guide
When NOT to implement a suggestion immediately
- Design questions: "Should this be a separate class?" — Present to user for decision
- Optional nitpicks: Reviewer marked as
nit:— Ask user if they want to include it - Large refactors: If fix would require changing >50 lines, propose a separate follow-up issue instead
When to ask the user before proceeding
- Any fix involving behavioral changes to plugin logic
- Renaming Valve keys (breaking change — requires migration notes)
- Changes that affect the bilingual release notes already committed
Anti-Patterns to Avoid
- ❌ Do NOT
git commit --amendon a pushed commit without user approval for force-push - ❌ Do NOT silently skip a reviewer's comment; always acknowledge it (implement or explain why not)
- ❌ Do NOT use
--force(only--force-with-leasewhen absolutely necessary) - ❌ Do NOT make unrelated changes in the fixup commit; keep scope focused on review feedback
- ❌ Do NOT respond to reviewer comments in Chinese if the PR language context is English
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

