Friday Router

What This Skill Does

IntentRouter, also known as Friday Router, acts as the intellectual traffic controller for your OpenClaw agent ecosystem. In a complex AI environment, the main orchestrator (typically Gemini 2.5 Flash) is tasked with managing high-level logic, but it should not handle specialized work like complex coding, creative writing, or deep research tasks. IntentRouter solves this by analyzing the semantic intent of your user inputs and mapping them to the specific Large Language Model (LLM) best suited for that objective.

By utilizing OpenRouter as its backbone, this skill allows you to leverage industry-leading models like MiniMax for coding, Kimi for creative endeavors, and Grok for rapid research. Version 1.7.0 represents a significant security pivot, focusing on a hardened, minimalist architecture. It has been stripped of gateway management and internal credential handling to ensure your sensitive tokens remain isolated. It acts as a pure decision-making node, returning structured JSON parameters that tell the orchestrator which specific session to spawn and which model to assign to that session.

Installation

To integrate Friday Router into your environment, use the OpenClaw command-line interface. Ensure your system has the proper permissions to create new session directories in your workspace.

clawhub install openclaw/skills/skills/runeweaverstudios/friday-router

Post-installation, ensure your OpenClaw instance is configured with a valid OpenRouter API key. Because IntentRouter uses the openrouter/ prefix for all model IDs, the router will automatically handle the routing logic without needing additional configuration files beyond the local config.json inside the skill directory.

Use Cases

IntentRouter is essential for power users who demand specialized performance for different task types.

• Code Development: Route structural architecture and debugging tasks to high-reasoning models like MiniMax while keeping your main orchestrator free for task planning.
• Creative Writing: Route character-driven, stylistic, or prose-heavy requests to Kimi, which excels at nuance and creative expression.
• Research & Synthesis: When the agent needs up-to-date information or rapid data aggregation, routing to Grok or similar research-optimized models ensures higher quality outcomes.
• Automated Workflows: By standardizing the routing of tasks into isolated sessions, you prevent "context bloat" where one massive conversation thread becomes unmanageable and slows down your agent's response time.

Example Prompts

1. "Analyze this Python script for memory leaks and suggest a performance optimization strategy."
2. "Write a short story about an AI agent discovering the concept of freedom, written in a melancholic and reflective tone."
3. "Conduct a comparative analysis of the current market trends for edge computing in 2025 and summarize the top three potential risks."

Tips & Limitations

• Orchestration Workflow: You must use the sessions_spawn function immediately after receiving output from the router. The router provides the data; the session spawn executes it.
• Security First: Never attempt to pass API keys or secrets through the router output. This skill is explicitly designed to exclude sensitive data from its payload.
• Model Compatibility: The router relies on the openrouter/ naming convention. If a model is not available via OpenRouter, it cannot be defined in the routing table.
• Performance: For best results, use the default orchestrator (Gemini 2.5 Flash) as the primary hub; it provides the lowest latency for tool-calling, which is necessary for the router to operate efficiently without impacting the user experience.