model-brain

What This Skill Does

The model-brain skill serves as the central orchestration hub for message routing within the OpenClaw environment. It intelligently intercepts incoming tasks and determines the optimal path forward by evaluating the complexity, risk level, and nature of the request. Instead of relying on a single, expensive LLM for every interaction, model-brain applies a cost-aware decision matrix to ensure that simple, deterministic tasks are handled by lightweight scripts while sensitive financial operations or high-complexity architectural challenges are escalated to premium, high-reasoning models.

Installation

To integrate this skill into your local environment, execute the following command in your terminal:

clawhub install openclaw/skills/skills/aaigotchi/model-brain

Ensure that you have python3 configured in your path, as the core routing logic resides within scripts/route_message.py.

Use Cases

• Cost Optimization: Automatically route casual conversational prompts to the cheapest available model while reserving high-tier compute for complex logic.
• Risk Management: Enforce security protocols by requiring the highest tier models for wallet interactions, preventing simple errors in sensitive financial transactions.
• Workflow Automation: Streamline your aaigotchi development by automatically deciding whether to call a local deterministic script (like pet-me-master) or spawn a specialized reasoning agent.
• Infrastructure Routing: Route repository-level edits directly to specialized code-focused models to ensure better context retention and bug-free implementation.

Example Prompts

1. "Check the current treasury balance and let me know if we need to rebalance the portfolio before the market opens."
2. "Draft a new function to calculate the gas fees for our upcoming contract migration, then perform a quick security audit on the logic."
3. "Pet all my gotchis and verify that their hunger levels are within the optimal range for the next 4 hours."

Tips & Limitations

• Deterministic First: Always attempt to categorize your request under the zero-llm path if a native skill already exists. This is the fastest, safest, and cheapest method for common repetitive tasks.
• Routing as Advice: Remember that model-brain provides a routing recommendation. It does not possess inherent transaction authority; your wallet-side security logic should always independently verify any instructions passed by a selected model.
• Environment Scaling: As you add new models to the bankr library, update your local references/routing-table.md to ensure the brain remains aware of your expanded capabilities. Use the scripts/select_model.sh entrypoint for quick shell-based integrations to keep your scripts clean and readable.