What problem does it solve? Agent skills written as SKILL.md files often drift in quality: vague descriptions, missing trigger conditions, stale commands, or broken scripts. This Skill scores any skill against an 11-point criteria set plus an Agent Skills spec gate, producing a structured report with evidence and concrete fix suggestions. ## Core Features & Use Cases - Criteria-based scoring: Judges 11 aspects (goal clarity, description quality, conciseness, freshness, client compatibility, and more) with four verdicts: pass, needs improvement, not applicable, or undetermined. - Spec compliance gate: Validates frontmatter structure, name rules, and description length (via check_description_length.sh) before scoring. - Freshness verification: Runs shellcheck, shfmt, and ruff on bundled scripts and executes them with side-effect-free arguments. - Criteria evolution tracking: Detects unrecorded criteria changes, drafts CHANGELOG entries, and proposes new criteria from post-review decisions. - Use Case: Point it at a skills directory to batch-review every SKILL.md, then get a summary of spec violations, top problem areas, and outdated patterns to remove. ## Quick Start Review the skill at skills/git/git-commit and report which criteria need improvement.
