gmail-automation

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

• Third-party content exposure detected (high risk: 0.75). The skill’s runtime workflow uses Composio/Rube MCP Gmail tools (e.g., GMAIL_FETCH_EMAILS, GMAIL_FETCH_MESSAGE_BY_MESSAGE_ID, GMAIL_GET_ATTACHMENT) which can ingest email body text authored by other people (outsider senders/threads) into the agent context for searching/replying/labeling.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill requires connecting to Rube MCP at runtime via https://rube.app/mcp (used by RUBE_SEARCH_TOOLS / RUBE_MANAGE_CONNECTIONS) to fetch tool schemas and perform remote tool actions, meaning external content from that URL directly controls agent behavior and executes actions.