How does this tool help with token management?

The reviewer flags 'over-explanations' of concepts Claude already knows, helping you strip unnecessary text and save tokens for critical conversation context.

What is the Claude Code Skill Reviewer?

It is a specialized tool that evaluates SKILL.md files to ensure they follow Anthropic's structural guidelines and qualitative best practices for high-performance AI agents.

What is the Auto-PR mode?

Auto-PR mode allows the reviewer to fork a repository and submit additive improvements to SKILL.md files, streamlining the upgrade process for existing skill sets.

Can I use this for private or team-specific skills?

Yes, the tool includes Scope Profiles (Project, Personal, Team, Public) that adjust the audit strictness based on the skill's intended audience.

Why are 'Principle' checks better than just linting?

Structural linting only checks formatting; Principle checks ensure the AI understands the reasoning behind rules, which prevents failures on edge cases and improves generalization.

Claude Code Skill Reviewer

Name: Claude Code Skill Reviewer
Author: nyjin

bynyjin

0•

Herramientas para Desarrolladores

Audits and optimizes Claude Code skill files using a 24-point quality checklist based on Anthropic best practices.

Skill Reviewer is a specialized utility designed to elevate the quality of Claude Code skills (SKILL.md) through a rigorous two-tier evaluation process. It validates structural correctness—such as YAML frontmatter and path formatting—while simultaneously auditing principle quality to ensure rules include reasoning, maintain token efficiency, and avoid overfitting. By categorizing issues from Critical to Low, it helps developers build more reliable, context-aware agentic capabilities that generalize across real-world edge cases rather than just passing simple syntax checks.

Características Principales

01Dual-tier evaluation covering 12 structural and 12 qualitative checkpoints

020 GitHub stars

03Severity-ordered reporting with actionable fix recommendations

04Context-aware scope profiles for Project, Personal, Team, or Public skills

05Auto-PR workflow for additive skill improvements and linting

06Token-efficiency auditing to maximize Claude's context window

Casos de Uso

01Standardizing agent behaviors and documentation within engineering teams

02Validating public-facing skills before submission to community marketplaces

03Refactoring legacy skills to improve trigger accuracy and reasoning depth

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nyjin/skill-reviewer skill-reviewer

For use in Claude.ai and ChatGPT

Características Principales

01Dual-tier evaluation covering 12 structural and 12 qualitative checkpoints

020 GitHub stars

03Severity-ordered reporting with actionable fix recommendations

04Context-aware scope profiles for Project, Personal, Team, or Public skills

05Auto-PR workflow for additive skill improvements and linting

06Token-efficiency auditing to maximize Claude's context window

Casos de Uso

01Standardizing agent behaviors and documentation within engineering teams

02Validating public-facing skills before submission to community marketplaces

03Refactoring legacy skills to improve trigger accuracy and reasoning depth

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add nyjin/skill-reviewer skill-reviewer

For use in Claude.ai and ChatGPT