Code Review Skill

Two-phase review: First analyze and document findings, then apply approved fixes.

Preset Configuration:

•Focus: Code quality, correctness, best practices
•Quality Gate: No new findings for 1 iteration (review comprehensive)
•Reviewer Type: Senior Code Reviewer
•Max Iterations: 4

When to Use

•After completing implementation of a feature
•Before creating a pull request
•When refactoring significant code
•Post-merge quality verification

Required Information

•Target - code path or git range (e.g., src-tauri/src/calculations.rs or HEAD~3..HEAD)
•Reference - plan or spec the code implements

Phase 1: Review (Findings Only)

Step 1: Gather Code Context

If git range provided:

bash

git diff --stat {BASE}..{HEAD}
git diff {BASE}..{HEAD}

If file path provided:

•Read the file(s)
•Identify recent changes

Step 2: Run Tests (Baseline)

bash

npm run test:backend

Note test status for review context (pass/fail). Do NOT fix yet.

Step 3: Create Review Document

Create _code-review.md in project root or task folder:

markdown

# Code Review

**Target:** {TARGET}
**Reference:** {REFERENCE}
**Started:** YYYY-MM-DD
**Status:** In Progress
**Focus:** Quality, correctness, best practices

**Baseline Test Status:** [Pass / N failures]

## Iteration 1

### Findings

[To be filled by review agent]

Step 4: Execute Review Loop

For each iteration (max 4):

Spawn Review Agent:

code

Task tool (superpowers:code-reviewer):
  Use template from requesting-code-review/code-reviewer.md

  WHAT_WAS_IMPLEMENTED: {description from reference}
  PLAN_OR_REQUIREMENTS: {REFERENCE}
  BASE_SHA: {BASE or N/A}
  HEAD_SHA: {HEAD or current}

  IMPORTANT: Document findings only. Do not apply fixes.

Update Review Document: Append findings to _code-review.md:

markdown

## Iteration N

### New Findings
- [Critical] Issue description - `file:line` - Suggested fix
- [Important] Issue description - `file:line` - Suggested fix
- [Minor] Issue description - `file:line` - Suggested fix

### Test Gaps
[Any missing test coverage noted]

### Coverage Assessment
[Areas reviewed / Areas remaining]

Commit Review Only:

bash

git add _code-review.md
git commit -m "review(code): iteration N findings for {TARGET}"

Quality Gate: Exit when no new findings for 1 iteration (review is comprehensive).

Step 5: Finalize Review Document

Update _code-review.md with summary:

markdown

## Review Summary

**Status:** Ready for User Review
**Iterations:** N
**Total Findings:** X Critical, Y Important, Z Minor
**Test Status:** [Pass / Fail]

### All Findings (Consolidated)

#### Critical
1. [ ] Issue - `file:line` - Suggested fix

#### Important
1. [ ] Issue - `file:line` - Suggested fix

#### Minor
1. [ ] Issue - `file:line` - Suggested fix

### Test Gaps
- [ ] Missing test for X
- [ ] Edge case Y not covered

### Recommendation
[Ready to merge / Needs fixes / Major issues]

Commit:

bash

git add _code-review.md
git commit -m "review(code): complete findings for {TARGET}"

Step 6: Present Review for Approval

Inform user:

Code review complete.

Please review _code-review.md for findings.

After your review, let me know:

•Which findings to fix

•Which to skip (with reason)

•Any questions about findings

STOP and wait for user direction.

Phase 2: Apply Approved Fixes

Only proceed after user approval.

Step 7: Apply Approved Fixes

For each user-approved finding:

•Implement the fix
•Check the finding as addressed in _code-review.md: [x]

Step 8: Run Tests

bash

npm run test:backend

If tests fail:

•Diagnose the failure
•Fix or report to user for decision

Step 9: Commit Changes

bash

git add -A
git commit -m "fix: address code review findings

Addressed:
- [list of fixed issues]"

Step 10: Final Assessment

Update _code-review.md:

markdown

## Resolution

**Addressed:** N findings
**Skipped:** M findings (user decision)
**Test Status:** All passing
**Status:** Complete

### Applied Fixes
- Finding 1: [how resolved]
- Finding 2: [how resolved]

### Skipped Items
- Finding X: [user's reason]

Domain-Specific Checklist

Review should verify:

• Tests pass
• No obvious bugs
• Error handling present
• No security vulnerabilities (input validation, etc.)
• Code matches plan/spec
• No scope creep (extra features not requested)
• Follows project patterns (see CLAUDE.md)

Example

code

User: /code-review src-tauri/src/suggestions.rs against _tasks/19-feature/02-plan.md

Claude: [Executes Phase 1 - runs tests, creates _code-review.md, iterates until comprehensive]
Claude: Code review complete. Please review _code-review.md for findings.

User: Fix all Critical issues. Skip the Minor style suggestions.

Claude: [Executes Phase 2 - applies approved fixes, runs tests, commits]