A strategic task orchestration skill that routes work to sub-agents using cost-benefit analysis and tiered model selection.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install delegate
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install delegate using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Delegate?
The Delegate skill is a sophisticated orchestration layer designed for AI agents to manage complex workflows by spawning specialized sub-agents. It operates on a fundamental economic principle: only delegate when the cost of spawning is lower than the cost of performing the task directly. This skill is essential for maintaining efficiency within Openclaw Skills by ensuring that high-resource models are reserved for architecture and complex reasoning, while smaller models handle formatting, searching, and classification.
By implementing a rigorous model tiering system, Delegate allows agents to intelligently choose between Small (Haiku/GPT-4o-mini), Medium (Sonnet/GPT-4o), and Large (Opus/o1) tiers based on the specific requirements of the deliverable. This approach minimizes latency and operational costs while maximizing output quality through specialized verification protocols.
Delegate Use Cases
- Scaling complex software development by delegating unit test generation to sub-agents.
- Reducing costs by routing documentation and research tasks to high-speed, low-cost model tiers.
- Managing multi-agent workflows where a primary controller orchestrates several specialized workers.
- Implementing automated error recovery and model escalation for tasks that fail to meet quality thresholds.
How Delegate Works
- Perform a cost-benefit analysis to determine if delegation is more efficient than direct execution.
- Select the optimal model tier based on the technical complexity and reasoning requirements of the task.
- Construct a spawn checklist containing the specific task, model tier, required context, and expected output format.
- Execute the sub-agent and monitor for lifecycle events such as timeouts or formatting errors.
- Conduct a verification phase to validate the sub-agent's work through test execution, syntax checking, and file verification.
Delegate Setup
To integrate this skill into your Openclaw Skills environment, ensure your agent has access to tiered model providers. Configuration involves setting up the escalation logic found in the core rules.
# Standard setup for delegation templates
cp templates.md ./skills/delegate/templates.md
# Configure error recovery thresholds
openclaw config set delegate.retry_limit 3
Delegate Data Schema & Taxonomy
The Delegate skill utilizes a structured taxonomy for every sub-agent request to ensure consistency across Openclaw Skills.
| Attribute | Description | Required |
|---|---|---|
| TASK | A single, clear deliverable definition | Yes |
| MODEL | Explicit tier choice (Small, Medium, Large) | Yes |
| CONTEXT | Minimal necessary files and information | Yes |
| OUTPUT | Specific format instructions (e.g., JSON, File path) | Yes |
| DONE | The specific signal or state that indicates completion | Yes |
Delegate Advanced Features
- Intelligent Escalation: Automatically shifts tasks from Small to Medium or Large tiers if output quality is insufficient.
- Failure Recovery: Implements specific recovery patterns for timeouts, including a single retry before aborting.
- Strict Verification: Prevents 'false completions' by mandating that agents run tests and check for physical file existence.
- Context Pruning: Enhances performance by ensuring sub-agents only receive the specific information needed for their task rather than the full conversation history.
SKILL.md
Loading
Related Openclaw Skills
Decide
A self-learning decision framework that enables AI agents to safely automate consequential project choices over time.
Database
A comprehensive operational framework for designing and maintaining high-performance databases while avoiding reliability and integrity pitfalls.
Dates
A privacy-first personal dating system built to track connections, plan thoughtful dates, and log meaningful reflections using Markdown files.
Database Manager
A robust database operating system for managing schemas, migrations, and recovery readiness with high reliability.
Deno
A comprehensive agent skill for building secure TypeScript applications using the Deno runtime while avoiding common permission and dependency pitfalls.
Deploy
A comprehensive framework for shipping applications reliably using modern CI/CD, rollback strategies, and zero-downtime patterns.
Featured*
