skill-creator

To create a reusable Claude skill, initialize a new template, customize the mandatory SKILL.md with YAML frontmatter, and bundle optional resources like scripts, references, and assets. This ensures structured workflows and domain knowledge extensions are consistently packaged.

The best way to package modular AI workflows is using a deterministic lifecycle of initialization, editing, validation, and packaging. This approach ensures progressive disclosure and reusability by keeping context lean while enabling on-demand access to detailed references.

Yes, you need YAML frontmatter in SKILL.md to build a valid skill package. The structured body and deterministic validation workflow require this frontmatter to properly define the skill's capabilities, metadata, and modular components for consistent execution.

Yes, you can bundle custom scripts and references into a Claude skill package. The skill structure explicitly supports optional resource folders for scripts, references, and assets, allowing you to integrate tooling and detailed documentation alongside the main SKILL.md file.

Progressive disclosure keeps context lean by structuring the skill to load only the core SKILL.md initially, while enabling on-demand access to detailed references, scripts, and assets. This pattern prevents context bloat while still providing deep domain knowledge and tooling.