English | 한국어
High-quality reviewer skills for Claude Code that evaluate and improve skills and custom commands against Anthropic's official best practices.
Evaluates Claude Code skills on structural correctness (F1-F12) and principle quality (P1-P12).
Use when:
- Creating a new skill
- Reviewing existing skills before publishing
- Contributing improvements to skill repositories
- Validating third-party skills before installation
Modes:
- Mode 1: Self-Review — validate your own skill
- Mode 2: External Review — evaluate someone else's skill
- Mode 3: Auto-PR — fork, improve, and submit PRs
- Mode 4: Create — generate new skills following best practices
Evaluates Claude Code custom slash commands on structural correctness (CF1-CF10) and principle quality (CP1-CP12).
Use when:
- Creating custom slash commands
- Reviewing
.claude/commands/*.mdfiles - Deciding between command vs skill format
- Improving command security and reliability
Modes:
- Mode 1: Self-Review — validate your command
- Mode 2: External Review — evaluate someone else's command
- Mode 3: Create — generate new commands following best practices
- Mode 4: Auto-PR — contribute improvements to command repos
Install directly from Smithery:
# Clone this repository
git clone https://github.com/nyjin/skill-reviewer.git
# Copy to your Claude skills directory
cp -r skill-reviewer/skills/* ~/.claude/skills/# Install only skill-reviewer
cp -r skill-reviewer/skills/skill-reviewer ~/.claude/skills/
# Or install only command-reviewer
cp -r skill-reviewer/skills/command-reviewer ~/.claude/skills/# Self-review your skill
/skill-reviewer
# Review a specific skill
/skill-reviewer Review the skill at ~/.claude/skills/my-skill
# Create a new skill
/skill-reviewer Create a skill for PDF processing
# Self-review your command
/command-reviewer
# Review a specific command
/command-reviewer Review the command at .claude/commands/deploy.md
# Create a new command
/command-reviewer Create a command for git commits
Form (Structural):
- F1-F4: Frontmatter (name, description, triggers)
- F5-F11: Body structure (line count, paths, references)
- F12: Script error handling
Principle (Quality):
- P1: WHY explanations
- P2: Token efficiency
- P3: Freedom calibration
- P4: Consistent terminology
- P5: Concrete examples
- P6: Feedback loops
- P7: Generalization
- P8-P12: Metadata quality
Form (Structural):
- CF1-CF6: Frontmatter (filename, YAML, description, allowed-tools)
- CF7-CF10: Security and portability
Principle (Quality):
- CP1: WHY explanations
- CP2: Token efficiency
- CP3: Freedom calibration
- CP4: Inline context (
!and@) - CP5: Output format
- CP6: Feedback loops
- CP7: Generalization
- CP8: Error handling
- CP9: Argument design
- CP10: Security (least privilege)
- CP11-CP12: Existence justification
Both reviewers adapt checks based on scope:
- 📁 Project: Single repo, overfitting acceptable
- 🏠 Personal: Your own use, must generalize
- 👥 Team: Shared within org, + documentation
- 🌐 Public: Open-source, all checks mandatory
Improvements welcome! Both reviewer skills follow their own best practices. When contributing:
- Run the reviewers on your changes
- Follow F/P or CF/CP checks
- Use the PR templates in
references/pull-request-guide.md
MIT License - see LICENSE