pre-pr-review

Pre-PR Review Skill

This skill provides automated local CodeRabbit reviews before creating pull requests, with systematic fix application and safety validation.

When This Skill Activates

Use this skill when you detect the user wants a local CodeRabbit review. Common trigger phrases:

• "Run local coderabbit scan"
• "Do a local coderabbit review"
• "Local CodeRabbit review"
• "CodeRabbit scan before PR"
• "Review my code locally"
• "Run coderabbit locally"

Key indicators:

1. Mentions "coderabbit" or "code rabbit"
2. Mentions "local", "locally", or "before PR"
3. Does NOT include a PR number (that would trigger coderabbit-workflow skill)

What This Skill Does

When activated, immediately delegate to the local-coderabbit-reviewer sub-agent by using the Task tool:

Use the Task tool with:
- subagent_type: "local-coderabbit-reviewer"
- description: "Run local CodeRabbit review"
- prompt: "Run a local CodeRabbit review against main branch. Scan all changes (committed + uncommitted), analyze results, present a plan for fixes, apply changes, run tests, and commit only if all tests pass."

Why Delegate to Sub-agent?

The local-coderabbit-reviewer sub-agent is specialized for this workflow and will:

1. Run CodeRabbit CLI scan (--plain --base main --config CLAUDE.md)
2. Save results to .coderabbit/review.txt
3. Read and analyze all findings
4. Present a plan for implementing fixes (even if not in plan mode)
5. Apply fixes systematically
6. Run full test suite
7. Only commit if tests pass ✅
8. Stop if tests fail (no commit) ❌

User Experience

Before this skill:

User: "Run local coderabbit scan"
Claude: [Manually runs command, asks what to do with results, may skip steps]

With this skill:

User: "Run local coderabbit scan"
Skill: [Auto-activates]
Sub-agent: [Scans → Analyzes → Plans → Fixes → Tests → Commits]
Result: ✅ Done automatically with plan presentation and safety checks

Safety Features

The local-coderabbit-reviewer sub-agent includes critical safety checks:

• Never commits failing tests - Stops and reports if any test fails
• Always presents plan - Shows what will be fixed before making changes
• Respects project patterns - Follows BCL conventions and standards
• Comprehensive scan - Reviews all changes (committed + uncommitted)

Differences from coderabbit-workflow

Supporting Files

This skill directory includes:

• SKILL.md (this file) - Skill definition and activation logic
• README.md - Detailed workflow documentation and examples

Configuration

No configuration required. The skill uses:

• CodeRabbit CLI: coderabbit (must be installed and authenticated)
• Sub-agent: local-coderabbit-reviewer (must exist in ~/.claude/agents/)
• Output directory: .coderabbit/ (created if doesn't exist)
• Base branch: main (configurable in sub-agent if needed)

Troubleshooting

If the skill doesn't activate:

1. Check that the user's message includes "coderabbit" + "local" keywords
2. Ensure there's no PR number (that triggers coderabbit-workflow instead)
3. Verify the local-coderabbit-reviewer sub-agent exists
4. Confirm CodeRabbit CLI is installed (coderabbit --version)

If CodeRabbit CLI is not installed:

The sub-agent will detect this and provide installation instructions:

curl -fsSL https://cli.coderabbit.ai/install.sh | sh
coderabbit auth login

If tests fail:

• The sub-agent will NOT commit changes
• Review test failures and fix them manually, or ask the user for guidance
• Changes remain in working directory for manual review

Example Interactions

Success Case

User: "Run local coderabbit scan"

Agent:
1. Running CodeRabbit scan against main branch...
   → Saving results to .coderabbit/review.txt

2. Analyzing 12 findings...

3. PLAN: Here's what I'll fix:
   • src/Domain/User.cs:45 - Add null check for email parameter
   • src/Application/UserService.cs:78 - Extract magic number 100 to constant
   • tests/UserTests.cs:120 - Improve test assertion message
   [... 9 more items ...]

   Proceed with fixes? [User confirms or plan mode auto-proceeds]

4. Applying fixes systematically...
   ✓ Fixed 12 issues

5. Running test suite...
   → All tests passed ✅

6. Committing changes...
   → Committed: "Address CodeRabbit review feedback"

✅ Local CodeRabbit review complete! Ready to create PR.

Failure Case (Tests Fail)

User: "Run local coderabbit scan"

Agent:
1. Running CodeRabbit scan...
   → Found 5 issues

2. Presenting plan...
   [Shows what will be fixed]

3. Applying fixes...
   ✓ Fixed 5 issues

4. Running test suite...
   ❌ 2 tests failed:

   UserServiceTests.CreateUser_WithInvalidEmail_ThrowsException
   UserServiceTests.UpdateUser_WhenNotFound_ReturnsError

🛑 Tests failed. Changes NOT committed.

Please review the failures. The fixes are in your working directory.

Best Practices

1. Run before every PR - Catch issues early before GitHub review
2. Review the plan - Understand what will change before it happens
3. Trust the safety checks - If tests fail, there's a reason
4. Keep .coderabbit/ local - Already in .gitignore, no need to commit
5. Use timestamped files - For history: .coderabbit/review-$(date).txt