---

name: agents-tasks-tools description: "Build and debug Langroid ChatAgent, Task, and ToolMessage workflows, including routing, orchestration tools, done sequences, structured output, batch helpers, XML tools, and MockLM tests." disable-model-invocation: true metadata: disco-role: operating license: MIT

Agents, Tasks, and Tools

Use this sub-skill for Langroid work centered on:

• ChatAgentConfig and ChatAgent
• TaskConfig and Task
• ToolMessage and XMLToolMessage
• orchestration tools such as DoneTool, ResultTool, and FinalResultTool
• TaskTool and RecipientTool
• done_sequences, handle_llm_no_tool, structured output, batch helpers, and MockLM

Use it when the problem is about agent orchestration, tool wiring, routing, termination, typed output, or deterministic testing. If the issue is about provider endpoints, keys, or model backends, route elsewhere first.

Route elsewhere

• Provider endpoints, keys, or model wiring → llm-provider-config
• RAG, document chat, or retrieval flows → retrieval-doc-chat
• SQL, table, or graph workflows → data-sql-graph-agents
• MCP or Chainlit integrations → integrations-mcp-chainlit

Core pattern

1. Pick the right ToolMessage or output type.
2. Enable it with enable_message.
3. Add a ChatAgent handler or tool-level handle / response method.
4. Wrap the agent in a Task.
5. Add routing or termination only when needed.
6. Test the flow with MockLM before using a provider-backed model.

Read next

• API reference
• Workflows
• Testing patterns
• Troubleshooting