Do I need to provide the full file path to the skill?

No, you only need to provide the skill name. The auditor uses a discovery strategy to locate all related files across your hooks, agents, and skills folders.

What happens if my skill doesn't use hooks?

The auditor intelligently detects the absence of hooks and will skip the Hooks Best Practice agent (Agent C) unless you explicitly request it.

Can I run just one part of the audit?

Yes, you can use flags like --structure-only, --content-only, or --hooks-only to trigger specific review agents instead of the full parallel suite.

What files does the review-skill analyze?

It automatically scans SKILL.md files, templates, agent definitions, hook scripts (PreToolUse, PostToolUse, etc.), settings.json, and associated test files in your .claude directory.

How does the parallel review work?

The skill spawns three specialized sub-agents—plugin-dev, skill-content-reviewer, and claude-code-guide—in a single message to evaluate different aspects of your skill simultaneously.

Skill Auditor & Reviewer

Name: Skill Auditor & Reviewer
Author: peterstorm

bypeterstorm

0•

Security & Testing

Audits and validates Claude Code skills using a multi-agent parallel review process to ensure structural integrity and content quality.

The review-skill is a specialized meta-utility designed to maintain high standards for Claude Code extensions by performing comprehensive 360-degree audits. It orchestrates three specialized agents—a structural reviewer, a content quality analyst, and a hooks best-practice guide—to evaluate skill manifests, agent definitions, and hook implementations in parallel. This tool is essential for developers who need to debug complex hook interactions, verify plugin compliance, or polish the documentation and logic of their custom skills before deployment.

Key Features

010 GitHub stars

02Hooks Best-Practice Validation: Checks registration formats, input/output contracts, and state management patterns.

03Deep File Discovery: Automatically crawls the .claude directory to find related agents, hooks, and test files.

04Unified Synthesis: Consolidates findings from multiple agents into a prioritized list of fixes and improvements.

05Parallel Multi-Agent Execution: Runs three specialized reviewers simultaneously for maximum efficiency.

06Granular Review Modes: Supports focused audits for structure, content, or hooks using specific command flags.

Use Cases

01Ensuring consistent documentation and prompt quality across a large library of custom agents.

02Debugging complex hook lifecycle events and state management issues within a new skill.

03Verifying that a third-party Claude Code plugin adheres to recommended structural conventions and security patterns.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add peterstorm/.dotfiles review-skill

For use in Claude.ai and ChatGPT

Download Skill