vocabulary: "1.0.0" info: provider: Agent Skills description: Vocabulary and taxonomy for the AI agent skills topic, covering the landscape of tools, capabilities, protocols, and frameworks that enable AI agents to perform tasks. created: '2026-04-19' modified: '2026-04-19' operational: apis: - name: Anthropic Tool Use API namespace: anthropic-tool-use version: "2023-06-01" baseUrl: https://api.anthropic.com status: active - name: Google Agent Development Kit namespace: google-adk version: "1.0" baseUrl: https://adk.dev status: active - name: Model Context Protocol namespace: mcp version: "1.0" baseUrl: https://modelcontextprotocol.io status: active resources: - name: tools description: Discrete skills or functions an agent can invoke apis: - anthropic-tool-use - google-adk - mcp actions: - define - invoke - discover - register - name: agents description: AI systems that orchestrate tool use and task completion apis: - google-adk - anthropic-tool-use actions: - create - run - orchestrate - compose - name: mcp-servers description: MCP protocol servers exposing skills to AI clients apis: - mcp actions: - build - deploy - connect - list - name: workflows description: Composed sequences of agent skills forming multi-step processes apis: - google-adk actions: - define - execute - monitor - name: sessions description: Stateful context sessions maintaining memory across agent interactions apis: - google-adk - anthropic-tool-use actions: - create - maintain - terminate actions: - name: invoke description: Execute an agent skill or tool httpMethod: POST pattern: write - name: define description: Declare a tool schema for agent use httpMethod: POST pattern: write - name: discover description: List available tools at runtime httpMethod: GET pattern: read - name: search description: Search available tools by capability httpMethod: POST pattern: query - name: execute description: Run a workflow or agent task httpMethod: POST pattern: write - name: compose description: Combine multiple skills into a multi-agent workflow httpMethod: POST pattern: write schemas: core: - name: Tool description: A skill definition including name, description, and input schema properties: - name: type - name: name - name: description - name: input_schema - name: ToolCall description: An agent's invocation of a tool with specific inputs properties: - name: id - name: type - name: name - name: input - name: ToolResult description: The output returned after executing a tool call properties: - name: tool_use_id - name: type - name: content - name: Agent description: An AI agent instance with associated skills and configuration properties: - name: name - name: model - name: tools - name: instructions - name: MCPServer description: A Model Context Protocol server exposing tools to AI clients properties: - name: name - name: transport - name: tools - name: resources parameters: identifiers: - name: tool_use_id description: Unique identifier for a tool invocation - name: session_id description: Identifier for a stateful agent session configuration: - name: model description: The AI model used by the agent - name: max_tokens description: Maximum tokens allowed in an agent response - name: tool_choice description: Controls how the agent selects tools (auto, any, tool, none) enums: tool_choice: - auto - any - none - tool transport: - http - stdio - websocket skill_type: - function - mcp - openapi - builtin builtin_skills: - web_search - code_execution - web_fetch - tool_search - file_read - file_write - bash - text_editor authentication: schemes: - name: API Key type: apiKey description: Bearer API key passed in x-api-key header apis: - anthropic-tool-use capability: workflows: - name: Tool Definition and Invocation description: Define agent skills as JSON schemas and invoke them from an AI model response apis: - anthropic-tool-use - google-adk personas: - AI Developer domains: - Tool Use - name: MCP Server Integration description: Connect an AI agent to an MCP server to discover and invoke remote skills apis: - mcp - anthropic-tool-use personas: - AI Developer - Platform Engineer domains: - MCP - Tool Use - name: Multi-Agent Orchestration description: Coordinate multiple specialized agents each with distinct skill sets to complete complex tasks apis: - google-adk personas: - AI Architect domains: - Multi-Agent - Orchestration personas: - id: ai-developer name: AI Developer description: Developers building AI-powered applications with tool use and agent capabilities workflows: - Tool Definition and Invocation - MCP Server Integration - id: platform-engineer name: Platform Engineer description: Engineers deploying and managing MCP servers and agent infrastructure workflows: - MCP Server Integration - id: ai-architect name: AI Architect description: Architects designing multi-agent systems with composed skill workflows workflows: - Multi-Agent Orchestration domains: - name: Tool Use description: The practice of enabling AI agents to invoke external functions and APIs - name: MCP description: The Model Context Protocol ecosystem for standardized agent-tool connectivity - name: Multi-Agent description: Systems where multiple AI agents collaborate, each contributing specialized skills - name: Orchestration description: Coordination of skill invocation sequences and agent delegation crossReference: - resource: tools operations: - define - invoke - discover workflows: - Tool Definition and Invocation - MCP Server Integration personas: - AI Developer - resource: mcp-servers operations: - build - deploy - connect workflows: - MCP Server Integration personas: - Platform Engineer - resource: agents operations: - create - run - orchestrate workflows: - Multi-Agent Orchestration - Tool Definition and Invocation personas: - AI Architect - AI Developer