deep-agents-orchestration
Agent workflowslangchain-ai/langchain-skillsskills.sh ↗
Installs
12,833
deduplicated, at the last sync
Since we started
+1.4%
40 readings, about 2 hours apart. Not a live curve.
Our category
Agent workflows
ours
Last read
Sep 3, 2026
from the directory
Our brief
oursThis skill orchestrates complex agent workflows by providing three core capabilities: delegating specialized tasks to subagents, planning multi-step operations using a todo list, and requiring human approval for sensitive actions. It allows the main agent to manage complexity, isolate work, and ensure compliance before executing critical steps.
- final report from subagent
- todo list of tasks with status
- state indicating pending action or decision
- user message/instruction
- subagents configuration (name, tools)
- thread_id for state persistence
not detected
The evidence does not provide information regarding paid plans or API costs.
not detected
The evidence does not mention any required registration process.
The skill cannot make subagents stateful, meaning they are ephemeral and require complete instructions in a single call. It also cannot configure tool names (like `task` or `write_todos`) or the HITL protocol structure itself. Furthermore, it requires a checkpointer when using human-in-the-loop interrupts.
Evidenceconfig/skills/deep-agents-orchestration/SKILL.md:1-438
--- name: deep-agents-orchestration description: "INVOKE THIS SKILL when using subagents, task planning, or human approval in Deep Agents. Covers SubAgentMiddleware, TodoList for planning, and HITL interrupts." --- <overview> Deep Agents include three orchestration capabilities: 1. **SubAgentMiddleware**: Delegate work via `task` tool to specialized agents 2. **TodoListMiddleware**: Plan and track tasks via `write_todos` tool 3. **HumanInTheLoopMiddleware**: Require approval before sensitive operations All three are automatically included in `create_deep_agent()`. </overview> --- ## Subagents (Task Delegation) <when-to-use-subagents> | Use Subagents When | Use Main Agent When | |-------------------|-------------------| | Task needs specialized tools | General-purpose tools sufficient | | Want to isolate complex work | Single-step operation | | Need clean context for main agent | Context bloat acceptable | </when-to-use-subagents> <how-subagents-work> Main agent has `task` tool -> creates fresh subagent -> subagent executes autonomously -> returns final report. **Default subagent**: "general-purpose" - automatically available with same tools/config as main agent. </how-sub
Read 1 of 1 text files in the skill.
Installfrom the directory
npx skills add https://github.com/langchain-ai/langchain-skillsInstalling happens there, not here. We are an index with an opinion, not a mirror.
What is inside itfrom the directory
1 files — names only. The directory does not report sizes.
What the auditors foundfrom the directory
A skill is instructions your agent will follow and scripts it may run, so who checked it matters as much as how many people installed it.
Installs, reading by readingours
Axis starts at 12.7k, not zero — the range is 12.7k to 12.8k.
Asked out loudspoken, not typed
The same skill in the words people use speaking to an assistant rather than typing into a box. Each one carries the situation it came from, and each answer says only what the skill's own files support.
Yes, you should use subagents when the task requires specialized tools or if you want to isolate complex work. This allows the main agent to delegate the job and receive a final report.
It includes TodoListMiddleware, which automatically creates a todo list for complex tasks. This helps track progress by setting items as pending, in_progress, or completed.
The skill supports HumanInTheLoopMiddleware. You can configure specific tools to require human approval before they execute, which is useful for high-stakes operations like deployments.