Skill Judge

This is a meta-skill that evaluates other skills against the official spec. It scores across dimensions like knowledge delta (does it teach Claude what it doesn't already know, or just waste tokens on basics?), mindset and procedures, and anti-pattern quality. The underlying philosophy is sharp: skills aren't tutorials, they're knowledge externalization mechanisms where good design means expert-only insights minus what Claude already has baked in. If you're building or reviewing skills and want to know whether you're actually adding value or just burning context window on "what is a for-loop" explanations, this gives you a rubric. The 120-point scoring system is thorough but opinionated, which is exactly what you want from a judge.

Install to Claude Code

npx -y skills add gohypergiant/agent-skills --skill skill-judge --agent claude-code