skill-creator

To create reusable Claude skills, you need a standardized scaffold with a SKILL.md file, validated frontmatter, and optional resource directories like scripts and references. This approach turns onboarding knowledge into self-contained, modular skill units while reducing duplication.

Scaffolding a new skill requires generating a template SKILL.md with required frontmatter and initializing directories for scripts, references, and assets. Using an automated initializer ensures consistent structure and encapsulates best practices for progressive disclosure and modular design.

Yes, pyyaml is required to parse and validate frontmatter when generating skill units. The dependency supports the validation scripts that check if your skill metadata meets the standardized structure requirements before packaging.

Yes, you can package modular skills using dedicated workflow scripts like package_skill.py. These scripts automate the validation of frontmatter and bundle the scaffolded directories into a distributable format.

Templates for skill creation provide a structural skeleton but require manual customization of the SKILL.md content and resources to fit specific use cases. The generated scaffold ensures correct frontmatter and directory structure, but does not write the actual logic.

Progressive disclosure in modular skills works by structuring the SKILL.md to expose core instructions while keeping detailed references and assets in separate directories. This design pattern ensures Claude loads only necessary context, optimizing performance.