What problem does it solve? Writing an Agent Skill that actually gets invoked is hard: descriptions that name tools instead of user needs never trigger, rules duplicated into CLAUDE.md make skills unreachable, and merges or removals silently orphan references. This Skill encodes the conventions, pitfalls, and verification gates needed to keep a repository of skills discoverable and conformant to the Agent Skills standard. ## Core Features & Use Cases - Skill lifecycle management: Create, edit, merge, absorb, or retire skills following naming conventions (nodefony-*), metadata.version placement, and progressive disclosure with references/ resources. - Compliance gate: The skills-doc.mjs script validates every skill against the Agent Skills standard (name format, description length, allowed frontmatter fields, dead references) and regenerates public documentation pages plus a machine-readable registry. - Trigger benchmark: The trigger-bench.mjs script measures whether real user phrases elect the right skill, catches over-broad descriptions with negative cases, and tracks trigger overlaps. - Script placement audit: The scripts-audit.mjs script classifies every repository script by whether it belongs in a skill, detects orphans and dead references, and distinguishes documented scripts from actually executed ones. - Use Case: A skill in your repository never gets invoked. Run the diagnostic procedure to check whether its rule is duplicated in CLAUDE.md, whether its triggers name the tool instead of the user need, and whether its references still resolve. ## Quick Start Ask the agent to audit why a specific skill never triggers, or to validate all skills in the repository against the Agent Skills standard before publishing.
