creating-skills

To create a modular AI agent skill, use the Skill Initializer to scaffold a new directory and customize the SKILL.md file with required name and description frontmatter to ensure proper triggering and packaging.

The best way to structure resources is by organizing them into scripts, references, and assets directories, which allows the AI agent to selectively load deterministic tasks, documentation, and output templates.

The skill validation process checks readiness for packaging by running verify.py to enforce frontmatter structure, validate naming conventions, and ensure all bundled resources are correctly organized.

Yes, you need a SKILL.md file containing valid frontmatter with a name and description to initialize the skill workflow, which then scaffolds the necessary scripts, references, and assets directories.

Yes, you can bundle optional resources into a self-contained skill by placing them in the scripts, references, and assets directories, allowing the agent to load specific components on demand.

Custom skills fail validation before packaging when the SKILL.md file lacks required frontmatter fields, uses invalid naming conventions, or contains unorganized resources that violate the modular skill structure.