skill-quality-reviewer

Point the reviewer at the skill directory containing SKILL.md and it runs an eight-step workflow covering frontmatter parsing, four scoring dimensions, and report generation. Output includes a weighted score, letter grade, and prioritized improvement plan.

A good description uses third-person phrasing like "This skill should be used when...", includes three to five specific trigger phrases, stays between 100 and 300 characters, and names concrete use cases rather than vague claims.

SKILL.md should stay under 5,000 words, with 1,500 to 2,000 words considered ideal. Detailed content belongs in a references/ directory so it loads only when needed, following progressive disclosure principles.

Yes, the batch-portfolio mode reviews multiple skills together, clusters repeated issues across the portfolio, and produces a prioritized shortlist of what to fix first using the batch review template.

Low structural scores usually come from missing name or description fields in the YAML frontmatter, invalid YAML syntax, or files referenced in SKILL.md that do not exist. The bundled skill-audit.py script detects missing references automatically.