What problem does it solve?
Switching or upgrading the model on an OpenClaw instance often fails in confusing ways: 401 "Invalid token" errors despite valid keys, "No available channel" errors, thinking-level rejections, or config edits that silently never take effect. This Skill diagnoses the real cause before editing and safely switches the model configuration.
Core Features & Use Cases
- Safe model switching: A Python script discovers all openclaw.json config locations, backs them up, adds known model definitions (k3, k2p6, kimi-k2.7-code), updates the default model, syncs mirror configs, and restarts the gateway.
- Diagnosis-first troubleshooting: A trap catalog covers env-var API key hijacking, provider-plugin thinking-level restrictions, relay group/network availability, and wrong-file edits, with probe commands to prove each cause.
- End-to-end verification: Mandatory real agent-turn verification checks for fallback usage so a masked failure is never mistaken for success.
- Use Case: Your OpenClaw bot falls back or errors on LLM calls after you tried upgrading from k2p6 to k3. The Skill probes the endpoint, finds that an env KIMI_API_KEY is hijacking the wire key, fixes the config, and verifies a clean agent turn.
Quick Start
Ask the assistant to switch the OpenClaw default model to k3 and restart the gateway, then verify the change with a real agent turn.
