| name | skill-center-curator |
| description | Use whenever working in 0langas-skill-center or another personal skill collection to add, rename, move, publish, document, test, or install skills. This skill orchestrates official-docs-citation-cache, tri-client-skill-port, skill-trigger-evalsmith, and agent-install-sync so every skill is canonical, documented, tri-client compatible, linked correctly, covered by activation evals, and safe to publish. |
| when_to_use | When adding, renaming, moving, publishing, testing, organizing, documenting, or installing skills in 0langas-skill-center; when README tables, provider links, evals, docs evidence, or local installs need to stay in sync; when the user says this repo is becoming a skill center. |
| type | prompt |
| whenToUse | When adding, renaming, moving, publishing, testing, organizing, documenting, or installing skills in 0langas-skill-center; when README tables, provider links, evals, docs evidence, or local installs need to stay in sync; when the user says this repo is becoming a skill center. |
| disableModelInvocation | false |
Skill Center Curator
Use this skill as the operating manual for 0langas-skill-center. It keeps the collection useful as it grows instead of letting it become a pile of disconnected skills.
Owned Outcome
After a skill-center task, the repo should have:
- one canonical top-level directory per skill
- tri-client frontmatter for Codex, Claude Code, and Kimi Code where appropriate
- provider discovery links under
.codex/skills, .claude/skills, and .kimi-code/skills
- trigger evals beside each skill
- official docs evidence cached when provider behavior is claimed
- README skill table updated
- local install state synced when requested
- tests passing
Orchestration Order
Use the other skills in this order:
official-docs-citation-cache for provider docs evidence.
tri-client-skill-port for metadata and canonical layout.
skill-trigger-evalsmith for activation evals.
agent-install-sync for provider and user-level links.
- Curator final pass for README, tests, commit readiness, and release notes.
Curator Workflow
- Inspect git status and existing skill directories.
- Identify the requested skill operation: add, rename, import, update, publish, or install.
- Run the orchestration order above.
- Update
README.md skills table and remove completed candidates from the candidate list.
- Run deterministic checks:
- frontmatter has required fields
- provider links exist and resolve
- eval JSON parses
- README table contains the skill
- official docs cache exists when provider claims changed
- Summarize only material changes and verification results.
Repository Policy
- Keep skills ASCII unless the source skill already uses Unicode intentionally.
- Do not duplicate skill content across provider folders.
- Do not delete user-level installs without explicit confirmation.
- Keep candidate ideas separate from implemented skills.
- Prefer high-value workflow-specific skills over generic commodity skills.
- If a skill would be better as a plugin, say so and stop before overloading the skill surface.
Output
{
"action": "skill-center-curation",
"skills_changed": ["skill-name"],
"docs_cached": true,
"tri_client_ported": true,
"evals_created": true,
"links_synced": true,
"tests": "passed",
"ready_to_commit": true
}