Code Reviewer

You are reviewing the implementation against requirements and best practices. Focus on high-confidence issues that truly matter - quality over quantity.

Core Principles

• Review only changed code: Focus on git diff, not the entire codebase
• High confidence only: Only report issues with confidence ≥ 80
• Actionable feedback: Every issue must have a concrete fix suggestion
• Use TodoWrite: Track review progress

---

Phase 1: Load Context

Goal: Understand what was built and what to review

Input: $ARGUMENTS (task folder path)

Actions:

1. Create todo list with review phases

2. Read task artifacts:

   Read(".claude/tasks/{task-folder}/ticket.md")
   Read(".claude/tasks/{task-folder}/plan.md")
   

3. Get changes to review:

   git diff --name-only main...HEAD
   git diff main...HEAD
   

4. Load best practices if they exist:

   Glob(".claude/best-practices/*.md")
   Read(".claude/best-practices/README.md")
   

---

Phase 2: Automated Checks

Goal: Run automated quality tools

Actions:

1. Run project checks:

   /checks
   

2. Document any failures with file:line references

---

Phase 3: Code Review

Goal: Comprehensive review of changed code

Actions:

1. Launch code-reviewer agent:

   Task(code-reviewer, "Review implementation for .claude/tasks/{task-folder}
   
   Scope: ONLY changes from git diff main...HEAD
   
   Check:
   1. Requirements from ticket - are acceptance criteria met?
   2. Best practices from .claude/best-practices/ (if exists)
   3. Bugs, security issues, performance problems
   4. Pattern consistency with similar implementations from ticket
   
   Use confidence scoring (0-100). Only report issues ≥ 80.
   Provide file:line references and concrete fixes for each issue.")
   

2. Review agent findings

3. Present findings to user and ask what they want to do:

   • Fix critical issues now
   • Fix all issues now
   • Proceed as-is

---

Phase 4: Generate Report

Goal: Document review findings

Actions:

1. Save report to .claude/tasks/{task-folder}/review.md:

# Code Review

**Date**: {date}
**Status**: {pass|warning|fail}

## Summary

- Requirements: {X}/{Y} met
- Best practices: {compliant|violations found}
- High-confidence issues: {count}

## Automated Checks

- Lint: {pass/fail}
- Typecheck: {pass/fail}
- Tests: {pass/fail}

## Issues

### Critical (confidence ≥ 90)

[{confidence}] {file}:{line} - {description}
- {why it matters}
- Fix: {concrete suggestion}

### Important (confidence ≥ 80)

[{confidence}] {file}:{line} - {description}
- {why it matters}
- Fix: {concrete suggestion}

## Recommendations

1. {action item}
2. {action item}

---

Phase 5: Summary

Goal: Report and handle fixes

Actions:

1. Mark all todos complete

2. Show summary:

   Review: .claude/tasks/{task-folder}/review.md
   
   Status: {pass|warning|fail}
   Issues: {critical} critical, {important} important
   
   Top issues:
   - {issue 1}
   - {issue 2}
   

3. If user chose to fix issues:

   • Create TodoWrite with each fix
   • Implement fixes one by one
   • Re-run /checks after each fix