vocabulary: "1.0.0" info: provider: Agent Skills description: Vocabulary for the Agent Skills topic, covering vendor-published skill repositories, skill definitions, manifests, runtimes, and governance for file-based AI agent capability packages. created: '2026-05-19' modified: '2026-05-19' operational: apis: - name: Anthropic Claude Skills namespace: anthropic-claude status: active - name: Resend Skills namespace: resend status: active - name: ngrok Agent Skills namespace: ngrok status: active - name: Postman Skills namespace: postman status: active - name: JFrog Skills namespace: jfrog status: active - name: Airbyte Claude Skills namespace: airbyte status: active - name: Apollo GraphQL Skills namespace: apollo-graphql status: active - name: Airtable Skills namespace: airtable status: active resources: - name: skills description: Individual skill bundles, each described by a SKILL.md and supporting assets actions: - list - get - install - activate - execute - deactivate - name: manifests description: Repository-level manifests listing all skills published by a vendor actions: - list - get - validate - name: scripts description: Executable assets (bash, python, node) shipped inside a skill bundle actions: - list - get - execute - name: bindings description: Mappings between a skill and a specific agent runtime, environment, or secret actions: - create - get - list - delete - name: policies description: Governance rules that gate which skills agents may activate actions: - list - get - apply actions: - name: list description: Enumerate skills, manifests, scripts, or policies httpMethod: GET pattern: read - name: get description: Retrieve a single skill, manifest, or related resource httpMethod: GET pattern: read - name: install description: Add a skill bundle to an agent's skill directory httpMethod: POST pattern: write - name: activate description: Load a skill into an active agent session so it can be invoked httpMethod: POST pattern: write - name: execute description: Run a skill or one of its scripts on behalf of the agent httpMethod: POST pattern: write - name: deactivate description: Unload a skill from an agent session httpMethod: POST pattern: write - name: validate description: Validate a skill or manifest against the schema and policy rules httpMethod: POST pattern: write - name: apply description: Apply a governance policy to a skill or skill bundle httpMethod: POST pattern: write schemas: core: - name: SkillDefinition description: "A single Agent Skill: a file-based capability package that an AI coding agent can discover, load, and execute" properties: - name - display_name - description - version - runtime - provider - repository - entrypoint - instructions_file - scripts - tags - license - trigger_phrases - required_secrets - name: SkillManifest description: Repository-level manifest listing the skills published by a vendor or organization properties: - manifest_version - name - provider - homepage - repository - runtimes - skills - policies - maintainers - tags enums: runtimes: - claude-code - claude-agent-sdk - cursor - openai-apps-sdk - gemini - generic script_languages: - bash - python - node - typescript - other skill_statuses: - draft - active - deprecated - removed capability: workflows: - name: Publish Vendor Skill Bundle description: A vendor authors, validates, and publishes a *-skills repository so AI coding agents can install their capabilities apis: - anthropic-claude - resend - postman personas: - Vendor DevRel Engineer domains: - Skill Authoring - name: Install and Activate Skill description: A developer or agent installs a vendor skill bundle and activates specific skills inside a Claude Code or Claude Agent SDK session apis: - anthropic-claude - ngrok personas: - Agent Developer domains: - Skill Runtime - name: Govern Skill Activation description: A platform team curates an allow-list of approved skills, signs manifests, and enforces policy at activation time apis: - anthropic-claude personas: - Platform Engineer domains: - Skill Governance - name: Generate Skills from OpenAPI description: Convert an OpenAPI specification into a skill bundle, one skill per operation, so an agent gets governed access to every endpoint apis: - apollo-graphql - postman personas: - API Producer domains: - Skill Authoring personas: - id: vendor-devrel-engineer name: Vendor DevRel Engineer description: Engineers at API vendors who design, author, and ship official skill bundles for their products workflows: - Publish Vendor Skill Bundle - Generate Skills from OpenAPI - id: agent-developer name: Agent Developer description: Developers building applications and workflows on top of Claude Code, the Claude Agent SDK, or similar agent runtimes workflows: - Install and Activate Skill - id: platform-engineer name: Platform Engineer description: Platform and security engineers governing which skills are allowed to run inside their organization's agents workflows: - Govern Skill Activation - id: api-producer name: API Producer description: API teams that ship OpenAPI/AsyncAPI specs and want their endpoints exposed as agent-callable skills workflows: - Generate Skills from OpenAPI domains: - name: Skill Authoring description: Authoring, validating, and publishing skill bundles, including SKILL.md, scripts, and assets - name: Skill Runtime description: Installing, activating, executing, and deactivating skills inside an AI agent runtime - name: Skill Governance description: Allow-listing, signing, auditing, and policy enforcement for skill activation across an organization crossReference: - resource: skills operations: - list - install - activate - execute workflows: - Install and Activate Skill personas: - Agent Developer - resource: manifests operations: - list - get - validate workflows: - Publish Vendor Skill Bundle personas: - Vendor DevRel Engineer - resource: policies operations: - list - apply workflows: - Govern Skill Activation personas: - Platform Engineer