Chitin Core

What This Skill Does

Chitin Core is a sophisticated intelligent routing engine for the OpenClaw ecosystem, designed to optimize model selection for every task assigned to your agent. By dynamically evaluating task requirements against available computational costs and performance capabilities, it ensures your sub-agents are spawned using the most cost-effective and capable model for the job. It acts as a middleware between your user requests and the underlying LLM providers, effectively preventing crashes during rate-limiting scenarios and enforcing logical resource allocation across three distinct capability tiers: LIGHT, MEDIUM, and HEAVY. It maintains system health through active telemetry and automated fallback mechanisms, ensuring that even if your primary models are under load, your agent remains operational.

Installation

To install Chitin Core, run the following command in your OpenClaw CLI: clawhub install openclaw/skills/skills/adroidian/chitin-core Ensure you have Node.js installed in your environment, as the skill relies on internal scripts within the workspace to route tasks and manage configurations. Once installed, verify the setup by running node ~/.openclaw/workspace/skills/chitin-core/scripts/router.js validate.

Use Cases

Chitin Core is essential for developers or power users running multiple agents. Use it to delegate complex coding tasks to HEAVY models while keeping simple status checks and conversational responses on cost-efficient LIGHT models. It is particularly useful for enterprise deployments where cost optimization is required and service stability is paramount. Use the override tags like @light or @heavy to manually tune the performance profile for specific task chains.

Example Prompts

1. "Delegate the refactoring of the authentication module to a sub-agent using @heavy mode."
2. "Route this: summarize the contents of the last five chat logs and tell me if there are any urgent issues."
3. "Spawn a sub-agent to analyze the attached data file; route this to the most efficient model available."

Tips & Limitations

Always check your config.json after updates to ensure your preferred provider models are correctly mapped to tiers. If you encounter consistent rate limits, do not manually force heavy models—instead, rely on the router to handle escalation. Note that while the router is highly intelligent, it performs best when task descriptions provided in the route command are descriptive. The skill is designed for autonomous fallback, meaning it will automatically degrade to local Ollama instances if external APIs are exhausted.