gemini-deep-research

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

• Third-party content exposure detected (high risk: 0.75). High: the required runtime workflow sends the Deep Research agent’s output text (outputs[-1].get("text") in polling mode or SSE content.delta/thought_summary deltas) into the client’s context/handling, and that output is generated from web-sourced, outsider-authored content the agent retrieves at runtime.