| name | check-skill-name |
| description | Checks proposed skill names for collisions with reserved or existing names. Use when validating a skill or command name before creation. |
Check Skill Name
Core Contract
Validate skill/command names before creation to avoid collisions across harnesses and this repo.
known-names.json is the source of truth; only update it in explicit update mode.
Follow repo naming rules from CLAUDE.md / AGENTS.md.
Required Inputs
- Proposed name(s).
- Mode:
check (default) or update.
- Current repo skill/agent names for in-repo collision checks.
Workflow
Check mode
- Normalize candidate names (lowercase, strip
/, basename only).
- Load reserved names from
known-names.json and in-repo names from frontmatter.
- Classify each candidate:
CONFLICT: exact reserved/repo match.RISKY: near-match or alias collision.CLEAR: no collision.
- For non-clear results, propose 3-5 safe alternatives.
- Report verdict plus list
updated date.
Update mode
- Read
sources in known-names.json.
- Fetch each source docs page and extract command/skill names.
- Merge names conservatively (do not delete uncertain misses).
- Rewrite
known-names.json with merged names and new updated date.
- Report per-source added/removed/unchanged and fetch failures.
Safety Rules
- Never create the skill directory or any skill files. This skill only validates a name; creating the skill is a separate step the user takes after a clear verdict.
- The only file this skill may write is
known-names.json, and only in update mode. Do not edit it during a check.
- In update mode, never blank out a source on a failed fetch; leave its existing names intact and report the failure.
- Do not invent names as "reserved" without a source; every reserved name traces to
known-names.json.
Output Style
For check: normalized name, verdict (CLEAR/CONFLICT/RISKY), owning sources, safe alternatives when needed, and updated date.
For update: per-source add/remove/unchanged counts, new updated date, and any failed refreshes.
