outlook-automation

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

• Third-party content exposure detected (high risk: 0.80). This skill explicitly reads and ingests user-generated email content and attachments via the Core Workflows (e.g., OUTLOOK_SEARCH_MESSAGES, OUTLOOK_GET_MESSAGE, OUTLOOK_LIST_OUTLOOK_ATTACHMENTS, OUTLOOK_DOWNLOAD_OUTLOOK_ATTACHMENT in the "Search and Filter Emails" and related sections of SKILL.md), so untrusted third-party message content could be interpreted and drive subsequent tool actions.

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 https://rube.app/mcp at runtime (via RUBE_SEARCH_TOOLS/RUBE_MANAGE_CONNECTIONS) to fetch tool schemas that directly determine agent instructions and available actions, so this external endpoint controls the agent's behavior.