Agent-Skills.md

Agent Skills: Code Review

Review pull requests, branch changes, or code diffs. Triggers on "review this PR", "review my changes", "code review", "review branch", or GitHub PR URLs. Focuses on bugs, tests, complexity, and performance - not linting.

UncategorizedID: tdhopper/dotfiles2.0/reviewing-code

Repository

tdhopper/dotfiles2.0
tdhopper
73

Install this agent skill to your local

pnpm dlx add-skill https://github.com/tdhopper/dotfiles2.0/tree/HEAD/.claude/skills/reviewing-code

Skill Files

Browse the full folder contents for reviewing-code.

Download Skill

Loading file tree…

.claude/skills/reviewing-code/SKILL.md

Skill Metadata

Name
reviewing-code
Description
Review pull requests, branch changes, or code diffs. Triggers on "review this PR", "review my changes", "code review", "review branch", or GitHub PR URLs. Focuses on bugs, tests, complexity, and performance - not linting.

Code Review

Focus on substantive issues: bugs, missing tests, complexity, performance, duplication, incomplete implementations. Skip linting concerns (formatting, imports, naming style).

External Review (Optional)

Check for external reviewers and use if available. Priority: Codex > Gemini

command -v codex >/dev/null 2>&1 && echo "Codex available"
command -v gemini >/dev/null 2>&1 && echo "Gemini available"

If Codex available:

  • Local branch: codex --config model_reasoning_effort="high" review --base BASE_BRANCH
  • Remote PR: gh pr diff NUMBER | codex review --config model_reasoning_effort="high" -

If only Gemini: Pipe diff to gemini with review prompt.

Workflow

  1. Get the diff

    • PR: gh pr view NUMBER --json title,body,files then gh pr diff NUMBER
    • Branch: git diff origin/master...HEAD
    • Uncommitted: git diff

  2. Gather context - Read PR description, commit messages, project CLAUDE.md

  3. Review each file for:

    • Completeness: All code paths handled? Stubs left behind?
    • Tests: Added? Meaningful? Edge cases covered?
    • Complexity: Justified abstractions? Simpler alternatives?
    • Performance: Hot path regressions? Unbatched I/O?
    • Duplication: Similar code already exists? (rg "pattern")

  4. Synthesize external review (if used) with your findings. Consensus issues = high confidence.

Output Format

## Summary
[1-2 sentences]

## External Reviewer
[If used: Codex or Gemini]

## Key Findings

### Must Address
1. **[Issue]** (`file:line`) [Models]
   - Details
   - **Risk**: Why it matters

### Should Consider
2. **[Issue]** (`file:line`)
   - Details

### Minor Notes
- Observations

## Tests
[Coverage and quality assessment]

## Complexity
[Net impact on codebase complexity]

Numbering: Single sequence across all sections. Model attribution: [Codex + Claude], [Claude], etc.

Scope

In scope: Logic errors, missing error handling, test gaps, performance regressions, unnecessary complexity, duplication, incomplete implementations, project guideline violations.

Out of scope (linters handle): Formatting, import order, naming style, type annotations, docstring format.