create-skill

To create reusable AI agent skills, use a structured workflow that gathers requirements, scaffolds directories, fills SKILL.md frontmatter, and validates the package before deployment. This ensures consistent quality and reduces duplication across teams.

Scaffolding a new agent skill creates a default directory structure including a SKILL.md file for instructions and frontmatter, alongside optional folders for scripts, references, and assets to bundle resources directly with the skill.

Yes, the skill creation workflow requires specific frontmatter in the SKILL.md file, mandating at minimum a name and description to define the skill's identity and scope before validation and packaging can proceed.

You can update existing AI agent skills by editing their SKILL.md operational guidance, adding new bundled resources, and re-running the validation and packaging workflows to prepare the updated version for reuse.

The best way to validate an AI agent skill before distribution is to run a dedicated validation script against the skill's directory to check frontmatter requirements and structural integrity prior to packaging.

You need to run a pre-creation interview to accurately gather requirements and define the skill's scope, ensuring the scaffolding, resources, and SKILL.md content precisely match the intended workflow before initialization.