<agents_system priority="1"> Agents are specialized AI assistants that run in independent contexts for complex multi-step tasks.
How to use agents (spawned with independent context):
- The from the agent entry contains the agent definition file
- The agent definition will be automatically loaded into the subagent's context
- Invoke: Task(subagent_type="", prompt="task description")
- The agent runs in a separate context and returns results
- Example: Task(subagent_type="code-reviewer", prompt="Review the authentication code in auth.ts")
Usage notes:
- Agents have independent context windows
- Each agent invocation is stateless
- Agents are spawned as subprocesses via the Task tool
- The agent's AGENT.md file is loaded into the subagent's context automatically
<available_agents>
nango-docs-migrator Migrates Nango integration documentation from old tabbed format to new simplified structure with separate guide pages and pre-built syncs sections .claude/agents/nango-docs-migrator.md