agent-orchestrate
Multi-agent orchestration patterns for OpenClaw. Quick reference for spawning sub-agents, parallel work, and basic coordination. Use when: simple parallel tasks, fan-out/fan-in, basic pipelines. For advanced dynamic orchestration (agent-built task trees, spawn vs fork, human-in-the-loop), see cord-trees skill instead.
Why use this skill?
Master multi-agent coordination with the OpenClaw agent-orchestrate skill. Learn to build parallel tasks, sequential pipelines, and human-in-the-loop workflows.
Install via CLI (Recommended)
clawhub install openclaw/skills/skills/moltonbot000/agent-orchestrateWhat This Skill Does
The agent-orchestrate skill provides the essential building blocks for multi-agent coordination within the OpenClaw ecosystem. It is designed to manage complex workflows by allowing a primary agent to spawn, monitor, and control sub-agents. This skill simplifies the process of breaking down large, monolithic tasks into smaller, manageable units that can be executed in parallel or sequentially. By leveraging primitives like sessions_spawn and subagents steer, users can create sophisticated pipelines, parallel fan-out patterns, and human-in-the-loop checkpoints, significantly increasing the reliability and depth of automated tasks.
Installation
To integrate this orchestration capability into your OpenClaw environment, execute the following command in your terminal:
clawhub install openclaw/skills/skills/moltonbot000/agent-orchestrate
Use Cases
- Parallel Research: Spawn multiple agents to investigate different facets of a topic simultaneously to save time.
- Sequential Pipelines: Chain agents together where the output of a research agent feeds directly into an analysis agent, which then feeds a writing agent.
- Human-in-the-Loop Workflows: Create workflows that pause for human approval before proceeding to the next high-stakes step.
- Complex Task Decomposition: Break down massive project requirements into a hierarchy of sub-tasks, ensuring each piece is handled by a specialized sub-agent.
Example Prompts
- "Orchestrate a research task: spawn 3 agents to gather information on Cloudflare, AWS, and GCP pricing, then combine their results into a single comparison report."
- "Start a two-stage pipeline: first, have an agent draft a technical outline for a blog post, and once that is complete, notify me for approval before starting the full article."
- "Run a dependency tree for my project: ensure the data cleaning agent finishes before triggering the data visualization agent."
Tips & Limitations
- Isolation vs. Context: Use the 'Spawn' (Clean Slate) approach for independent tasks to maintain context clarity and reduce cost. Use 'Fork' (Context-Inheriting) only when subsequent tasks depend heavily on the specific data generated by previous sub-agents.
- Monitoring: Always use
subagents listto keep track of active sessions. Orphaned agents can continue to consume system tokens if not properly terminated. - Advanced Needs: If you find yourself needing to dynamically build task trees or manage complex state-based recursion, transition your workflow to the
cord-treesskill for more advanced logic. Keep your orchestration logic clean by modularizing each sub-agent's prompt for better observability.
Metadata
Not sure this is the right skill?
Describe what you want to build — we'll match you to the best skill from 16,000+ options.
Find the right skillPaste this into your clawhub.json to enable this plugin.
{
"plugins": {
"official-moltonbot000-agent-orchestrate": {
"enabled": true,
"auto_update": true
}
}
}Tags(AI)
Flags: code-execution, file-read, file-write