social-inbox-agent

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

• Third-party content exposure detected (high risk: 1.00). The agent explicitly fetches user-generated DMs and comments from the Late API (e.g., GET /v1/inbox/conversations and GET /v1/inbox/conversations/{id}/messages) and feeds that untrusted social-media content into the Gemini prompt to generate replies and drive sending actions, so third-party content can directly influence behavior.