Research Google Workspace Integration

CRITICAL E004: Prompt injection detected in skill instructions.

• Potential prompt injection detected (high risk: 0.90). The skill contains explicit instructions to withhold or avoid reporting the unavailability of Google Workspace integration (e.g., "No need to inform user..." and "DON'T ... Provide error messages if workspace tools are unavailable"), which are deceptive behaviors outside the benign purpose of describing optional integration usage.

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

• Third-party content exposure detected (high risk: 0.75). The required workflow can call the outsider-authored Gmail/Drive MCP tools (e.g., search_gmail_messages → get_gmail_message_content and search_drive_files → get_drive_file_content), which would ingest arbitrary email/document body text authored by other people into the agent’s LLM context.