Model Router
PassAudited by ClawScan on May 1, 2026.
Overview
This is a coherent instruction-only model router; the main thing to review is its automatic choice of model providers and pricing tiers.
This skill appears safe to install as an instruction-only routing guide, but set clear preferences for maximum model cost and avoid gateway auto-routing for sensitive prompts unless you trust the providers involved.
Findings (2)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
The agent may select cheaper or more premium models for tasks without asking each time, which can change cost and output quality.
The skill directs the agent to influence model/sub-agent execution choices. This is central to the stated model-routing purpose, but it can affect cost, quality, and execution behavior.
Before executing any task via `sessions_spawn` or delegating to a sub-agent, classify the task complexity using the rules below and route to the optimal model.
Confirm your preferred cost ceiling and ask the agent to get approval before using premium tiers for sensitive or expensive tasks.
Prompts may be handled by OpenRouter-selected third-party models rather than one explicitly chosen provider.
The skill recommends a gateway-style auto-routing option where the exact downstream model/provider may vary. This is expected for a model router, but it affects where prompt data may be processed.
Or use `openrouter/free` to auto-route across all free models.
Avoid auto-routing sensitive prompts unless you are comfortable with the gateway and downstream providers' data-handling policies.