Skip to main content

Skills and Tools

Tools are executable capabilities. Skills are instruction packages that teach an agent when and how to use capabilities.

Tools

Tools can read files, call APIs, search the web, query knowledge (RAG), manage documents and tasks, schedule jobs, drive a browser session, generate media, run code in the sandbox, and interact with connected integrations through MCP servers.

Tool access is governed by:

  • Agent configuration (the agent's bound tool set).
  • Runtime policy.
  • Workspace permissions.
  • HITL approval rules.

Two search-shaped tools are worth distinguishing: search_documents finds files by name and metadata; rag retrieves content evidence from inside documents. Agents are steered to the right one automatically, but knowing the split helps when reading execution logs.

Skills

A skill is a reusable recipe — "how to produce a weekly digest", "how to triage a support email" — that loads into an agent's context only when relevant, keeping prompts focused.

A skill usually contains a SKILL.md file (YAML frontmatter plus the instruction body) and optional assets or scripts. Scripts execute in the sandbox, not in the application processes.

Manage skills under Skills in the sidebar or via the API:

EndpointPurpose
GET /api/v1/skills / POST /api/v1/skillsList / create
POST /api/v1/skills/generateGenerate a skill from a description
POST /api/v1/skills/install-githubInstall a skill from a GitHub repository

Skills also power automations: when you schedule an agent job, Manor freezes the prompt into a generated execution skill so every run behaves consistently.

Writing Good Skills

  • Name the use case clearly.
  • Explain required inputs.
  • Describe tool usage boundaries.
  • Include examples for common workflows.
  • Avoid broad permissions when narrower tools are available.