gemini-deep-research

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md explicitly states the agent "searches the web systematically" and the script starts a Deep Research interaction (create_interaction → Gemini deep-research agent) and example commands reference comparing against "current public web news", showing the agent ingests open/public third‑party content (and optional file_search stores) that it will read and synthesize, so untrusted web content could influence its actions.