Review

Code Review

You are conducting a code review. Use a sub-agent to perform a thorough review of the current changes.

Instructions

1. First, determine what code to review:

   • If on a feature branch, review changes vs main/master
   • If specific files were mentioned, focus on those
   • Otherwise, review recently modified files

2. Spawn a sub-agent with subagent_type: Explore to review the code using the rubric below.

3. The sub-agent should return a structured review with:

   • Summary of changes
   • Issues found (categorized by severity)
   • Suggestions for improvement

Code Review Rubric

Have the sub-agent evaluate against these criteria:

Correctness (Critical)

• Logic errors or bugs
• Edge cases not handled
• Race conditions or concurrency issues
• Resource leaks (files, connections, memory)

Security (Critical)

• Input validation missing
• Injection vulnerabilities (SQL, command, etc.)
• Secrets or credentials in code
• Improper authentication/authorization

Error Handling (High)

• Exceptions not caught or handled appropriately
• Error messages that leak implementation details
• Missing error recovery or cleanup

Code Quality (Medium)

• Functions too long (>50 lines)
• Deep nesting (>3 levels)
• Code duplication
• Poor naming (unclear variable/function names)
• Missing or incorrect type hints

Testing (Medium)

• New code lacking tests
• Tests not covering edge cases
• Tests that are flaky or order-dependent

Documentation (Low)

• Public APIs without docstrings
• Complex logic without explanatory comments
• Outdated comments

Output Format

Present the review as:

## Review Summary
[Brief description of what was reviewed]

## Issues Found

### Critical
- [Issue description] (file:line)

### High
- [Issue description] (file:line)

### Medium
- [Issue description] (file:line)

### Low
- [Issue description] (file:line)

## Suggestions
- [Optional improvements that aren't issues]

## Verdict
[APPROVE / REQUEST_CHANGES / NEEDS_DISCUSSION]