box-automation

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

• Third-party content exposure detected (high risk: 0.80). The skill’s workflows (see "Search and Browse Content" and the Upload/Download steps in SKILL.md) explicitly call BOX_SEARCH_FOR_CONTENT and BOX_DOWNLOAD_FILE and list content_types like "file_content", "comments", and "web_link", meaning it fetches and reads user-generated/untrusted third-party content from Box items which can be used to determine subsequent actions (IDs, moves, deletes, shares), enabling indirect prompt-injection risk.

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 the Rube MCP server at https://rube.app/mcp at runtime (via RUBE_SEARCH_TOOLS) to fetch tool schemas that directly determine the agent's tool-invocation instructions, making this an external runtime dependency that can control prompts.