skills.rest
ricardoo022
ricardoo022COMMUNITY
Explore all skills by this author →
GitHub

deep-agents-orchestration

Configure subagent delegation, todo planning, and human approval interrupts in Deep Agents.

Updated May 1, 2026
Search Tags:#langchain#subagents#agent orchestration#task delegation#human-in-the-loop#deep agents#todo planning
One-click install
npx skills add https://github.com/ricardoo022/4dill --skill deep-agents-orchestration-ricardoo022
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: deep-agents-orchestration
Source: https://github.com/ricardoo022/4dill/tree/main/.gemini/skills/deep-agents-orchestration
Command: npx skills add https://github.com/ricardoo022/4dill --skill deep-agents-orchestration-ricardoo022

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Building multi-agent workflows with LangChain Deep Agents requires correctly wiring subagent delegation, task planning, and human-in-the-loop approval, and misconfiguring any of these (missing checkpointers, lost thread IDs, stateless subagent assumptions) causes silent failures. ## Core Features & Use Cases - SubAgentMiddleware: Delegate work to specialized subagents via the task tool, with custom tools, prompts, and per-subagent interrupt rules. - TodoListMiddleware: Plan and track multi-step tasks with the write_todos tool using pending/in_progress/completed statuses. - HumanInTheLoopMiddleware: Pause execution before sensitive tool calls and approve, reject, or edit proposed actions via Command(resume=...). - Use Case: Build a code-deployment agent where a subagent runs tests and deploys, but the deploy_to_prod tool requires human approval before execution, with the full workflow resumable via a thread ID. ## Quick Start Create a deep agent with a custom subagent and interrupt_on approval for a sensitive tool, then resume the paused run with an approve decision using the same thread ID.

Frequently Asked Questions about deep-agents-orchestration

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I create a custom subagent in Deep Agents?▼

Pass a subagents list to create_deep_agent with each entry defining name, description, system_prompt, and tools. The main agent then delegates work through the built-in task tool, and the subagent executes autonomously and returns a final report.

How do I add human approval before a tool runs in LangChain agents?▼

Set interrupt_on in create_deep_agent mapping tool names to True or an allowed_decisions list, and provide a checkpointer such as MemorySaver. When the tool is proposed, execution pauses and you resume with Command(resume={"decisions": [{"type": "approve"}]}).

Why does my Deep Agents interrupt fail or lose state?▼

Interrupts require both a checkpointer and a consistent thread_id in the invoke config. Without a checkpointer the agent cannot pause, and without the same thread_id the interrupted workflow cannot be resumed.

Are Deep Agents subagents stateful across multiple task calls?▼

No, subagents are stateless and start fresh on every task call. Provide complete instructions in a single call, including where to save outputs and what to return, rather than relying on follow-up calls.

Do custom subagents inherit skills from the main agent?▼

No, custom subagents do not inherit the main agent's skills; you must pass a skills list explicitly in each subagent's configuration. Only the default general-purpose subagent inherits the main agent's configuration.

Related Skills

Explore More Software Engineering Skills→
langchain-ailangchain-ai
1.1k

LangChain Dependencies

Guide installation and version management for LangChain ecosystem packages.

Official
Intermediate
jxtngxjxtngx
2

langchain-middleware

Implement human-in-the-loop approval and custom middleware for LangChain agents.

Community
Advanced
Reofdev07Reofdev07

deep-agents-orchestration

Coordinate subagents, TodoList planning, and human approvals in Deep Agents workflows.

Community
Advanced
brivarobrivaro

langgraph-fundamentals

Guide building stateful agentic workflows with LangGraph StateGraph nodes and edges.

Community
Intermediate