write-a-skill

To create an agent skill, generate a self-contained skill directory with a SKILL.md entry point containing discovery-ready YAML frontmatter for the name and description, plus concise instruction content.

Progressive disclosure in skill authoring keeps the SKILL.md entry point concise while linking to deeper reference materials. Use it when instruction complexity grows and supplementary documentation is needed.

Structure an agent skill package by writing a valid SKILL.md entry point and organizing optional supporting files for detailed documentation, references, or examples to ensure maintainability.

Deterministic scripts are optional in an agent skill package. You should add them only when necessary for specific task requirements, keeping the skill maintainable and concise.

The best way to organize references for maintainable agent workflows is splitting large instructions into additional reference materials within a self-contained skill directory, keeping the main SKILL.md concise.

Yes, skill authoring applies when designing capabilities for a specific task or domain. It prompts for the domain, use cases, required scripts, and references to scope the skill correctly.