Setup & Installation
Or with the ClawHub CLI, for registry-managed skill folders outside a full OpenClaw workspace:
Version History
v3.0.1: Sanitized config — private provider IDs removed, config.json.example added for clean public setup. 15-dimension task scoring, auto-model discovery, tier-based routing (SIMPLE/MEDIUM/COMPLEX/REASONING/CRITICAL), 80-95% cost savings on cron/subagent tasks.
What This Skill Does
Classifies tasks into five complexity tiers and assigns the cheapest model capable of handling each one. Sub-agents and cron jobs that skip model selection default to expensive Sonnet, so this skill enforces explicit routing before they are created. Classification uses a 15-dimension scoring system covering reasoning markers, code presence, multi-step patterns, token count, and constraint density.
Without explicit routing, every automated task defaults to Sonnet, and enforcing tier-based assignment at creation time cuts costs 80-95% on routine workloads.
When to use it
- Routing server monitoring cron jobs to free local models
- Preventing sub-agents from defaulting to expensive Sonnet
- Auditing existing cron payloads for missing or blocked model assignments
- Auto-discovering which provider models are currently live and responsive
- Delegating security reviews to Opus while sending heartbeats to cheaper models
Example Workflow
Here's how your AI assistant might use this skill in practice.
User asks: route a server health-check cron job to the cheapest available model
- 1Runs `router_policy.py recommend "check server health"` to classify the task
- 2Receives SIMPLE tier result with recommended model `anthropic-proxy-6/glm-4.7`
- 3Constructs cron payload including the recommended model ID
- 4Validates payload with `router_policy.py check` before submitting
- 5Creates the cron job with the cheap model instead of Sonnet
Cron job created with anthropic-proxy-6/glm-4.7 at $0.50/M instead of claude-sonnet-4-6 at $3/M
Requirements
Accounts, API keys, or tools you or your AI assistant may need to set up while using this skill.