dropbox-automation

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md explicitly instructs the agent to read and act on file content from Dropbox and external public sources (e.g., DROPBOX_READ_FILE, DROPBOX_GET_SHARED_LINK_FILE, and DROPBOX_SAVE_URL), which are untrusted/user-generated third‑party contents and are used to verify results and drive subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly requires connecting at runtime to the external MCP endpoint https://rube.app/mcp to obtain Rube MCP tool schemas and to run toolkit actions, so fetched content from that URL controls the agent's available instructions and executes remote tool behavior.