INVOKE THIS SKILL when using subagents, task planning, or human approval in Deep Agents. Covers SubAgentMiddleware, TodoList for planning, and HITL interrupts.
The deep-agents-orchestration skill is designed to enhance the capabilities of Deep Agents by providing orchestration features that facilitate task delegation, planning, and human approval processes. This skill addresses the challenges faced when managing complex tasks that require specialized tools or human oversight, allowing users to streamline their workflows and improve efficiency. By leveraging subagents, users can delegate specific tasks to agents that are tailored for particular functions, ensuring that the main agent can focus on broader objectives without being bogged down by intricate details.
You can create a custom subagent by defining its name, description, system prompt, and the tools it will use when initializing the deep agent.
No, subagents are stateless. You must provide complete instructions in a single call.
Yes, the skill provides examples in both Python and TypeScript, making it compatible with those languages.
If a sensitive operation is requested, the HumanInTheLoopMiddleware will require approval before proceeding.
The documentation does not specify a limit, but practical constraints may apply based on your implementation.
Quick Setup:
.claude/skills/Repository
langchain-ai/langchain-skills