email-check

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches unread emails from external Gmail/Outlook via MCP (e.g., search_gmail_messages, get_gmail_messages_content_batch, get_gmail_thread_content, list-mail-messages, get-mail-message in "Fetch Unread Emails" / workflow steps) and parses subject/body to score, categorize, and draft replies, so arbitrary user-generated third-party email content can directly influence agent decisions and actions.