skill-creator

To create Claude Copilot skills, draft a SKILL.md file, assemble test prompts, and run evaluation loops to measure trigger accuracy and quality. This workflow guides the end-to-end lifecycle from initial drafting to deployment packaging.

A SKILL.md file defines a Claude skill using structured descriptions and frontmatter metadata. Frontmatter validation enforces required fields and formatting rules, ensuring the skill is discoverable and correctly packaged into a deployable artifact.

Evaluate skill trigger accuracy by running trigger evals against test prompts and comparing results with baselines. An optimization script then refines the skill description to improve performance across iterations.

Yes, you can coordinate optional bundles including scripts, references, and assets within your skill. Once frontmatter is validated, these components are packaged together into a deployable .skill artifact.

Yes, PyYAML is required because skill frontmatter relies on YAML format for metadata definition. You need the PyYAML dependency installed in your environment to parse and enforce frontmatter validation.

The best way to iterate on a skill description is to run eval loops comparing quality and performance metrics against baselines, then use the improve_description script to refine the text for better trigger accuracy.