openrouter-model-routing

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill’s runtime routing/auto-routing functions read free-text prompts from callers (e.g., auto_route() extracts m["content"] from messages) and then send that text to OpenRouter via client.chat.completions.create, so an outsider can supply arbitrary prompt text that the workflow ingests.