Problem
Author documentation is strong, but creating a skill still requires manual directory, metadata, structure, reference, test, and eval-fixture setup that is easy to make inconsistent.
User outcome
An author can create a structurally complete skill and receive actionable local feedback before opening a pull request.
Scope
- Add an init command for canonical metadata and skill structure.
- Add a validate command covering metadata, references, adapter rendering, structure, catalog output, and eval fixture presence.
- Offer templates without inventing domain guidance.
- Document the review path for official and organization-specific skills.
Non-goals
- Generating substantive security advice with an LLM.
- Automatically accepting a skill into the official catalog.
Acceptance criteria
Validation
- Generate, complete, validate, render, install, and remove a fixture skill.
- Test malformed metadata, missing references, unsupported agents, and stale generated output.
Relationships
- Uses the machine-readable catalog schema.
- Feeds verifiable evaluation work.
Problem
Author documentation is strong, but creating a skill still requires manual directory, metadata, structure, reference, test, and eval-fixture setup that is easy to make inconsistent.
User outcome
An author can create a structurally complete skill and receive actionable local feedback before opening a pull request.
Scope
Non-goals
Acceptance criteria
Validation
Relationships