---
name: multi-model-routing-guard
description: Audit multi-model OpenClaw routing plans for fallback coverage, cost caps, tool-calling variance, unavailable model risk, and safe degradation before using Grok, Claude, OpenAI, Gemini, local models, or ACP harnesses together.
---
Multi-Model Routing Guard
Use when
Use when designing or validating OpenClaw model routing across multiple providers, fallback chains, cost tiers, tool-calling capabilities, or ACP harnesses.
Workflow
1. Provide a JSON routing file or directory containing model/task notes.
2. Run `python3 scripts/check_model_routing.py --target <path> --json`.
3. Fix missing fallbacks, undefined cost caps, and tool-call incompatible assignments before production use.
4. Report `SAFE_ROUTE`, `ROUTE_WITH_GUARDS`, or `ROUTE_BLOCKED`.
Parameters
→ `--target`: JSON file or folder with routing notes/configs.
→ `--json`: emit machine-readable output.
→ `--self-test`: run built-in fixture tests.
Output
Returns provider coverage, fallback gaps, cost-risk findings, and routing recommendation.
Release evidence
Reviewed in the 2026-06-15 GetAgentIQ warehouse propagation cycle.