coderabbit

name	coderabbit
description	Use when working with CodeRabbit AI code review. Covers triggering reviews, interpreting feedback, addressing comments, and PR workflow integration.

CodeRabbit Integration

Overview

CodeRabbit provides AI-powered code review for pull requests. This skill covers how to trigger reviews, interpret feedback, and address comments effectively.

When to Use

• PR created and awaiting review
• Need to understand CodeRabbit comments
• Want to re-trigger or customize review
• Integrating CodeRabbit into PR workflow

Quick Reference

Action	Command/Method
Trigger review	Auto on PR, or @coderabbitai review
Full review	@coderabbitai full review
Specific files	@coderabbitai review src/auth/*
Ignore file	@coderabbitai ignore on comment
Summary	@coderabbitai summary

---

Triggering CodeRabbit Review

Automatic Trigger

CodeRabbit reviews PRs automatically when:

• New PR is opened
• New commits are pushed to PR branch

Manual Triggers

# Full review of entire PR
@coderabbitai full review

# Review specific files
@coderabbitai review src/components/Button.tsx

# Review specific paths
@coderabbitai review src/auth/*

# Get summary only
@coderabbitai summary

# Generate PR description
@coderabbitai generate description

---

Interpreting Feedback

Comment Types

Type	Meaning	Action Required
Bug	Potential runtime error	Fix immediately
Security	Security vulnerability	Fix immediately
Performance	Inefficient code	Evaluate and fix
Best Practice	Code quality suggestion	Consider applying
Style	Formatting/convention	Apply if team standard
Nitpick	Minor suggestion	Optional

Priority Order

1. Security issues - Fix before merge
2. Bugs - Fix before merge
3. Performance - Evaluate impact, usually fix
4. Best practices - Apply when reasonable
5. Style/Nitpicks - Team discretion

---

Addressing Comments

Workflow

1. Read CodeRabbit comment
2. Determine action:
   - Fix: Make code change, push commit
   - Explain: Reply with reasoning
   - Disagree: Reply with justification
   - Ignore: Mark as resolved with reason

3. Reply to comment:
   - "Fixed in abc123"
   - "Intentional because [reason]"
   - "Won't fix: [justification]"

4. Resolve conversation when addressed

Response Templates

Fixed:

Fixed in [commit hash]. Added validation as suggested.

Intentional:

Intentional - this pattern is used for [reason].
See [link to docs/discussion] for context.

Won't Fix:

Acknowledged, but won't fix because:
- [reason 1]
- [reason 2]
Marking as resolved.

Need Clarification:

@coderabbitai Can you clarify? I don't understand how [X] applies here.

---

CodeRabbit Commands

In PR Comments

# Pause reviews on this PR
@coderabbitai pause

# Resume reviews
@coderabbitai resume

# Ignore specific file in future reviews
@coderabbitai ignore path/to/file.ts

# Get help
@coderabbitai help

Configuration

CodeRabbit is configured via .coderabbit.yaml:

reviews:
  auto_review: true
  path_instructions:
    - path: "src/components/**"
      instructions: "Focus on React best practices"
    - path: "src/api/**"
      instructions: "Check for proper error handling"

---

Common Patterns

Bulk Address Feedback

When CodeRabbit raises multiple similar issues:

# 1. List all comments
gh pr view --comments | grep -A5 "coderabbit"

# 2. Make bulk fix
# ... apply pattern fix across files ...

# 3. Single commit addressing all
git commit -m "fix: address CodeRabbit review feedback

- Add error handling to all API calls
- Fix type safety issues
- Update deprecated APIs"

# 4. Push and reply
git push

Disagree Professionally

Thanks for the suggestion. I've considered this but will keep
the current approach because:

1. [Technical reason]
2. [Team convention]
3. [Performance consideration]

Happy to discuss further if you see issues I'm missing.

---

Integration with PR Workflow

Pre-Merge Checklist

• All CodeRabbit security issues addressed
• All CodeRabbit bugs fixed
• Performance suggestions evaluated
• Unaddressed comments explained

When to Request Human Review

After CodeRabbit review:

1. Address all critical issues
2. Reply to suggestions
3. Push fixes
4. Then request human reviewers

---

Red Flags - STOP

Never:

• Ignore security findings
• Merge with unresolved bugs
• Dismiss without explanation

Always:

• Read each comment fully
• Respond to actionable items
• Document disagreements
• Fix security issues immediately

---

Integration

Related skills: submit-pr, git-expert Pairs with: verification, tdd