Review A Skill

Goal

Guardrails

• Always use

  --format json

  for machine-readable output in automated pipelines.
• Always use

  --fields

  to limit output size when inspecting catalog entries.
• Always use

  --dry-run

  before curating or removing a skill.
• Never remove a skill without first checking if other skills depend on it via

  info --format json

  dependencies.

Workflow

1. Preview the skill content to check for quality and safety.

npx ai-agent-skills preview <skill-name>

1. Inspect the catalog entry for metadata completeness.

npx ai-agent-skills info <skill-name> --format json --fields name,description,tags,collections,dependencies

1. Validate the skill's SKILL.md structure.

npx ai-agent-skills validate <skill-name>

1. If the skill needs curation (notes, collections, verification):

npx ai-agent-skills curate <skill-name> --notes "Reviewed: solid patterns" --verify --dry-run
npx ai-agent-skills curate <skill-name> --notes "Reviewed: solid patterns" --verify

1. If the skill should be removed:

npx ai-agent-skills curate <skill-name> --remove --dry-run
npx ai-agent-skills curate <skill-name> --remove --yes

Decision Criteria

• Keep: Clear description, valid frontmatter, useful to the library's audience, no injection patterns.
• Curate: Needs better whyHere, collection placement, or verification status.
• Remove: Duplicate, outdated, broken source, or contains suspicious content.

Gotchas

• The

  preview

  command only works for vendored (house) skills. Upstream skills show description and whyHere only.
• The

  validate

  command checks frontmatter structure but not content quality — that requires human or agent judgment.
• Removing a skill that other skills depend on will break the dependency graph. Always check

  dependencies.usedBy

  first.