skills-creator

To create an AI agent skill with valid YAML frontmatter, use skill scaffolding to generate a canonical SKILL.md file containing required name and description fields aligned with the agentskills.io specification.

Progressive disclosure design in agent skills enables AI agents to load only essential discovery metadata at startup, deferring full instruction loading until skill activation to optimize performance and context window usage.

You validate agent skills for deployment by applying validation rules against the skills-ref specification, checking required frontmatter fields like name and description, and ensuring structured organization of directories.

Yes, you can use existing reference directories when updating agent skills, as the scaffolding supports optional metadata and structured folders like scripts, references, and assets to maintain organization.

The best way to structure metadata for AI agents is using a canonical SKILL.md file with required frontmatter fields, recommended directory structures, and validation aligned to the agentskills.io specification.

Yes, aligning with the agentskills.io specification is required to scaffold a new skill, as it ensures consistent metadata, validates frontmatter fields, and enables structured guidance for AI agents.