open standard · S-043
Agent Skills
Package reusable instructions with optional references, assets, and scripts.
Markdown · JSON · Canonical source
- Revision
- living
- Maturity
- open standard
- Authority
- Agent Skills maintainers
- Structure
- SKILL.md with YAML frontmatter and directory conventions
- Roles
- skill-author, consuming-agent, host
An Agent Skill is a directory whose SKILL.md provides metadata and instructions. Optional references, assets, and scripts can extend the package.
The catalog separates discovery from progressive loading, resource resolution, script execution, and scope precedence. Recognizing a skill directory does not establish identical lifecycle or trust behavior across hosts.
Catalog mapping
Atomic capability rows
Absence here means the feature tree has not yet been published, not that the specification lacks the capability.
- Agent SkillsTrack every important Agent Skills behavior independently.
- Agent Skills core and discoveryRecognize, discover, and load reusable instruction packages centered on SKILL.md.
- Agent Skills manual invocationLet a user or agent explicitly select and activate a named skill.
- Agent Skills progressive loadingLoad skill metadata first and defer full instructions or resources until relevant.
- Agent Skills references and assetsResolve supporting reference files, templates, and assets from a skill package.
- Agent Skills scriptsResolve and execute optional scripts included with an Agent Skill under explicit runtime policy.
- Project-scoped Agent SkillsDiscover skills installed for a repository or workspace.
- User-scoped Agent SkillsDiscover personal or global skills outside the current project.
Conformance boundary
Proposed tests
Definitions only. No fixture has been executed against a harness.
- T-SKL-001SKILL.md frontmatter validation · required
- T-SKL-002Metadata-only startup discovery · recommended
- T-SKL-003Script execution relative to skill root · optional