Skill Library Maintainer

Maintain skill repositories as a product, not a dump of markdown files.

Workflow

1. Audit the collection structure before editing: identify priority folders, malformed metadata, duplicate coverage, and user-specific collections.
2. Fix high-value skills first; do not churn hundreds of files without a reason.
3. Keep

   SKILL.md

   lean and trigger-oriented: frontmatter should help discovery, body should explain workflow, and bulky details should move into

   references/

   .
4. Add

   agents/openai.yaml

   for curated skills that should surface cleanly in UI-driven environments.
5. Validate every touched skill after editing and record any repo-wide issues left unresolved.
6. Commit only the intentional paths and avoid sweeping unrelated repo cleanup.

Guardrails

• Do not preserve invalid frontmatter just because it already exists.
• Avoid redundant skills with nearly identical triggers unless they target distinct user groups.
• Prefer a few well-maintained flagship skills over many thin placeholders.
• Treat imported external collections as reference material until curated locally.

References

• Read

  references/audit-checklist.md

  before a broad update.
• Read

  references/curation-rules.md

  when deciding whether to rewrite, split, merge, or add a skill.