/rabbit - CodeRabbit AI Review

Purpose

Run CodeRabbit CLI for external AI code review. Provides a second opinion from an independent AI system with different analysis patterns.

Authentication (Recommended)

For enhanced reviews, authenticate first:

coderabbit auth login

Benefits of authentication:

• Team learnings applied to reviews
• Full contextual analysis
• Team standards enforcement
• Personalized recommendations

Check auth status in Error Handling section if issues occur.

When to Use

Use /rabbit	Use other command
Quick external review	Comprehensive review → /audit
Before commit/PR	Architecture review → /audit
Second opinion needed	Internal patterns check → /audit
Specific branch diff	Project rule compliance → /audit

Comparison with /audit

Aspect	/rabbit	/audit
Source	External AI (CodeRabbit)	Internal agents
Speed	Fast (10-30s)	Slower (multi-agent)
Focus	General code quality	Project-specific rules
Best for	Quick sanity check	Comprehensive review

Execution Time

Review time varies by scope:

Scope	Time	Recommendation
Small (< 10 files)	10-30 seconds	Run inline
Medium (10-50 files)	1-5 minutes	Run inline or background
Large (50+ files)	7-30+ minutes	Run in background

For large reviews, tell your AI agent to run CodeRabbit in the background and check back later.

AI Agent Best Practices

When using /rabbit with AI agents (Claude Code, Cursor, etc.):

1. Loop limits: Limit to 2-3 review cycles to prevent infinite fix loops
2. Background for large PRs: Use run_in_background: true for 50+ file changes
3. Don't chase perfection: Fix critical issues, accept minor suggestions as-is

Recommended workflow:
/rabbit → /fix (critical only) → /rabbit (verify) → /commit

NOT:
/rabbit → /fix → /rabbit → /fix → /rabbit → ... (infinite loop)

Usage

Basic (review uncommitted + recent commits)

/rabbit

Review against specific branch

/rabbit --base main

Review only uncommitted changes

/rabbit --type uncommitted

Review with custom config

/rabbit --config CLAUDE.md

Sandbox Requirement

CRITICAL: CodeRabbit CLI requires TTY/Raw mode for its terminal UI (Ink framework).

When executing coderabbit review, you MUST use dangerouslyDisableSandbox: true in the Bash tool call:

Bash({
  command: "coderabbit review --prompt-only",
  dangerouslyDisableSandbox: true  // Required for TTY support
})

Why: The sandbox environment provides non-TTY stdin, causing:

• "Raw mode is not supported" errors
• Hanging at "Connecting to review service"

Execution

Step 1: Check Git Status

!`git status --short | head -10`

Step 2: Run CodeRabbit Review

Execute CodeRabbit with appropriate options:

# Default: review all changes
coderabbit review --prompt-only

# With base branch (if specified)
coderabbit review --prompt-only --base <branch>

# With type filter (if specified)
coderabbit review --prompt-only --type <type>

Step 3: Parse and Display Results

Format the output for readability:

## 🐰 CodeRabbit Review Results

### Issues Found

| File | Line | Type | Description |
|------|------|------|-------------|
| ... | ... | ... | ... |

### Summary

- Total issues: X
- By type: refactor (N), potential_issue (N), ...

### Recommended Actions

1. [Priority issues to address]
2. [Optional improvements]

Step 4: Offer Next Steps

---
📋 Next steps:
- Apply fixes: `/fix` with specific issues
- Full review: `/audit` for comprehensive analysis
- Commit: `/commit` if issues are acceptable

Output Format

Minimal (no issues):

✅ CodeRabbit: No issues found

With issues:

🐰 CodeRabbit found N issues:

📁 path/to/file.ts
  L42 [refactor] Description of suggestion
  L78 [issue] Description of potential problem

📁 another/file.ts
  L15 [refactor] Another suggestion

💡 Run `/fix` to address these issues

Error Handling

Error	Action
Not a git repo	Display error, suggest git init
No changes to review	Display "Nothing to review"
Auth required	Suggest coderabbit auth login
Network error	Retry once, then display error

Integration with Workflow

Development Flow:
  /code → /rabbit → /fix (if needed) → /commit

PR Flow:
  /rabbit --base main → /audit → /pr