write-a-skill

To create a standardized AI agent skill, use a markdown template with required SKILL.md frontmatter containing name and description fields, plus optional directories for scripts, references, and assets to ensure structural consistency.

The best way to structure skill metadata is computing dependency, component, and toxicity fields within the frontmatter, while cataloging optional directories like scripts and assets to produce structured, reusable metadata.

The skill drafting workflow operates by guiding teams through a compact process: gathering requirements, drafting the SKILL.md content, and conducting a review phase to accelerate and standardize new skill creation.

You do not need optional directories to build a skill unit; they are entirely optional. A valid skill requires only a SKILL.md file with mandatory name and description frontmatter, while scripts, references, and assets enhance it.

Yes, you can use this approach to catalog dependencies and components for AI skills. The template computes and enforces structured metadata fields, detecting optional directories and organizing them into a standardized skill unit.

A key limitation is that the frontmatter strictly enforces mandatory name and description fields; without them, the skill unit is invalid. Additionally, it focuses on template generation and does not execute the generated scripts.