ai-model-web

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

• Third-party content exposure detected (high risk: 0.90). The "Custom onboarding (models outside the managed catalog)" section and its CreateAIModel example require registering an arbitrary BaseUrl (e.g. an OpenAI‑compatible endpoint) and then calling ai.createModel / generateText / streamText, which means the agent will fetch and consume responses from untrusted third‑party model endpoints that could contain instructions influencing subsequent behavior.