agents-sdk

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime path can ingest outsider-authored free text via chat/webhook/email inputs (e.g., AIChatAgent/saveMessages from client messages, routeAgentEmail raw email body, or webhook request.text()), which then becomes LLM-readable context for subsequent turns.