IntentRouter is an intelligent task director that analyzes user intent and delegates Openclaw Skills tasks to the most efficient LLM tier.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install friday-router
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 friday-router 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 IntentRouter?
IntentRouter serves as the central traffic controller for advanced AI workflows, specifically designed to enhance Openclaw Skills. By acting as a sophisticated orchestrator, it eliminates the inefficiency of using a single model for every task. Instead, it analyzes incoming requests and routes them to specialized providers via OpenRouter—leveraging MiniMax 2.5 for engineering, Kimi k2.5 for creative work, and Grok Fast for deep research.
Built with a security-first architecture, IntentRouter v1.7.0 removes sensitive gateway auth exposure, ensuring that session management remains isolated and secure. It operates as a mandatory delegation layer for orchestrator models like Gemini 2.5 Flash, ensuring that complex tasks are handled by sub-agents optimized for specific technical or creative domains.
IntentRouter Use Cases
- Delegating complex coding and debugging tasks to high-performance engineering models.
- Routing creative writing, UI/UX design, and frontend development requests to specialized creative LLMs.
- Automating research and web search tasks using high-speed reasoning models.
- Maintaining security by separating intent routing from gateway credential management in Openclaw Skills.
- Optimizing API costs by using fast, lightweight models for classification and powerful models only when necessary.
How IntentRouter Works
- The main orchestrator agent receives a user prompt and passes it to IntentRouter using the
spawn --jsoncommand. - IntentRouter analyzes the text against specific tier detection rules (Fast, Reasoning, Creative, Research, Code, Quality, Vision).
- The skill outputs a JSON object containing the refined task, the optimal OpenRouter model ID, and the session target.
- The orchestrator calls the
sessions_spawntool using the exact parameters provided by the router. - The specialized sub-agent executes the task and returns the result to the orchestrator.
- The orchestrator performs output hygiene to strip internal metadata and presents the clean response to the user.
IntentRouter Setup
To integrate IntentRouter into your Openclaw Skills environment, ensure you have an OpenRouter API key configured. Install the skill and use the following CLI pattern for routing:
# Navigate to the skill directory
cd workspace/skills/friday-router
# Run a test classification to verify routing logic
python3 scripts/router.py classify "Refactor this React component"
# Generate spawn parameters for your agent
python3 scripts/router.py spawn --json "Write a research paper on quantum computing"
IntentRouter Data Schema & Taxonomy
IntentRouter maintains a lean data footprint by relying on a local configuration file for its logic. It specifically avoids accessing global system secrets.
| Component | Description |
|---|---|
config.json |
Local file containing model mappings, tier keywords, and routing logic. |
router.py |
The core execution script for classification and spawning logic. |
| Spawn Output | A JSON object containing task, model, and sessionTarget. |
| Metadata | Internal routing excludes all gateway tokens and passwords for security. |
IntentRouter Advanced Features
- Multi-tier keyword detection including specialized triggers for React, JWT, and Vision tasks.
- Support for isolated session targets to prevent context contamination between sub-agents.
- Automated tier-bumping for agentic or multi-step tasks that require higher reasoning capabilities.
- Clean output hygiene filters that remove internal orchestration metadata before user delivery.
- Fail-safe delegation rules that prevent the orchestrator from attempting complex tasks it isn't optimized for.
SKILL.md
Loading
Related Openclaw Skills
Better Ralph
A structured autonomous coding agent that implements features story-by-story based on a project requirements document.
Agent Swarm
An intelligent orchestration skill that routes tasks to the most efficient AI models while managing sub-agent delegation.
Sessions Manager
A command-line utility for managing OpenClaw agent sessions, providing full visibility and control over session states and history files.
GitHub Projects Explorer
A powerful tool for searching and analyzing trending GitHub repositories using multi-dimensional filters like topics, stars, and creation dates.
Gateway Guard
A synchronization and recovery tool that maintains OpenClaw gateway authentication and automatically resumes agent sessions after errors.
RTL-SDR FM RDS Recorder
A specialized Linux toolset for high-quality FM radio recording and RDS station name extraction using RTL-SDR hardware.
Featured*
