Codex harness missing
Catch the exact failure before every fallback masks it.
Recovery Edition
OpenClaw update broke GPT? Fix the route, not just the symptom.
A field-tested kit for OpenClaw 5.12 operators dealing with GPT/Codex routing, plugin drift, dead fallbacks, and post-update smoke tests.
$10
Route drift
Understand when openai/gpt-* is safe and when it is not.
Bad fallback chain
Find stale Ollama or non-existent provider routes before they cascade.
Post-update proof
Run health, config, main chat, and subagent smoke checks.
What you get
- Update preflight checklist for capturing version, route, plugin, and rollback state before doctor touches config.
- GPT/Codex route guide explaining
openai/gpt-5.5,openai-codex/gpt-5.5, andagentRuntime.id=codex. - Read-only verifier script that checks gateway health, Codex package presence, active fallbacks, and Ollama contamination.
- Post-update smoke test for main chat, subagents, fallback safety, and desktop/client mismatch clues.
- Rollback checklist with clear criteria for when to restore config and when to keep the new route.
- Troubleshooting map from symptom to likely cause to fix.
Built from a real incident
This is not generic setup advice. It is the checklist we needed after a live GPT/Codex route failure.
No secret sharing
The verifier reads local config and prints route state without exposing API keys or tokens.
Audit upsell ready
If the kit shows a messy setup, the next step is a $297 Agent Audit.
Need us to inspect your setup?
Book an Agent Audit if you want a prioritized fix list for model routing, plugin state, cost leaks, and reliability risks.
Book an Agent Audit