Skip to main content

Agents

Agents packages role, prompt, model preference, skills, memories, and tool permissions into reusable execution profiles.

Open Agents when​

  • role quality matters more than one-off prompt wording
  • you keep repeating the same output rules or tool boundaries
  • you need to test or version a reusable assistant profile

Current surface​

  • built-in and custom agents
  • prompt, model, and skill editing
  • tool allow and deny policies
  • testing, duplication, import or export, and snapshots

Works closely with​

  • Chat for live execution
  • Skills for reusable domain behavior
  • Pipeline, Timer, and Channels for automation ownership

Deeper docs​

  • GitHub user handbook: docs/user/MODULE_PLAYBOOK_EN.md#5-agents
  • GitHub technical handbook: docs/technical/MODULE_ARCHITECTURE_EN.md#5-agents