An intelligent routing engine that dispatches specialized tasks to optimized AI models while maintaining a stable main session.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install advanced-dispatcher-skill
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 advanced-dispatcher-skill 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 Advanced Dispatcher?
Advanced Dispatcher is a high-performance management layer designed to orchestrate multi-model workflows within an AI agent environment. By analyzing the domain of a task—whether it be complex architecture, mathematical algorithms, or long-context research—it deterministically routes work to the most capable model available, such as GPT-5.4 or GLM-5. This Openclaw Skills extension ensures that developers can leverage the unique strengths of various LLMs without manually switching contexts.
The skill is built on a foundation of strict operational rules, emphasizing deterministic routing over ad hoc judgment. It enforces trunk-based development practices and integrates automated testing directly into the lifecycle of a task. With built-in safeguards, such as prompt-scoped model restrictions and specific cache retention policies, it provides a robust framework for professional-grade AI-assisted development.
Advanced Dispatcher Use Cases
- Executing structured multi-step development workflows using the buildq:, build:, or buildx: command prefixes.
- Performing automated code simplification and dead-code removal during mid-session refactoring.
- Evaluating technical tradeoffs by generating parallel proposals from different model architectures.
- Routing long-context research and documentation analysis to models optimized for high-token windows.
- Maintaining atomic, test-verified git diffs through automated branch management and PR-ready gates.
How Advanced Dispatcher Works
- The user initiates a request or a specific build pipeline command within the agent session.
- The dispatcher classifies the work domain (e.g., Code, Math, Web Dev) or identifies a specific build trigger.
- A deterministic routing table is consulted to select the optimal spawned model for the specific task.
- For coding tasks, the skill creates or verifies a short-lived feature branch to adhere to trunk-based development standards.
- The task proceeds through a sequence of pipeline steps including planning, implementation, and automated testing.
- A judge model validates the output against a strict contract, ensuring architecture and test plans are met.
- Results are integrated back into the main session only after passing all defined validation gates.
Advanced Dispatcher Setup
To deploy this skill within your Openclaw Skills environment, follow these configuration steps:
# Navigate to your skills directory
cd skills/advanced-dispatcher
# Ensure all dependencies for the dispatcher logic are installed
pip install -r requirements.txt
# Execute the validation suite to confirm routing logic and model connectivity
python test_dispatcher.py
After installation, ensure your environment variables are configured with the necessary API keys for the models defined in the routing table (OpenAI and OpenCode providers).
Advanced Dispatcher Data Schema & Taxonomy
The Advanced Dispatcher organizes task execution and model data using the following taxonomy:
| Data Type | Description | Retention Policy |
|---|---|---|
| RoutePlan | JSON object containing model choice, rationale, and pipeline steps | Per-request |
| Code Artifacts | File diffs and architectural plans generated by gpt-5.4 | Long cache |
| Analytical Data | Mathematical proofs and algorithmic logic from glm-5 | Short cache |
| Build Logs | Step-by-step execution results from build pipelines | Session-scoped |
| Judge Metadata | Risk lists, failure modes, and implementation checklists | PR-scoped |
Advanced Dispatcher Advanced Features
- Prompt-scoped model forcing allows for specialized tools like Claude to be utilized only when explicitly requested via the --force-claude flag.
- Tiered build pipelines (quick, standard, and strict) provide variable levels of rigor based on task complexity.
- Automated 'Simplify Contract' execution which removes speculative abstractions and over-engineered interfaces without user intervention.
- Multi-model tradeoff protocols that use a secondary judge model to evaluate and select the best architectural approach.
- Integrated cache retention management that optimizes performance based on the specific model provider used.
SKILL.md
Loading
Related Openclaw Skills
RDK X5 Media Support
A comprehensive multimedia toolkit for the RDK X5 platform enabling advanced audio, video, and display configuration.
Creator Monetization Risk Checker
A pre-publish compliance tool designed to audit video scripts and content for monetization risks, policy violations, and advertiser safety.
Tavily Research Pro
A professional-grade AI research engine providing structured information gathering and semantic analysis.
News Fact-Check Assistant
A comprehensive framework for verifying news claims, evaluating source reliability, and identifying misinformation through multi-source cross-referencing.
Short Video Hook Lab
A high-retention script architect that generates and scores 12 distinct hook variants to stop the scroll on TikTok, Reels, and Shorts.
Redbook Cards
An AI-powered content slicing tool that transforms long-form articles into high-engagement Xiaohongshu-style visual card sets in HTML format.
Featured*
