curated-automation

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

• Third-party content exposure detected (medium risk: 0.65). The required runtime workflow calls RUBE_SEARCH_TOOLS against the external Rube MCP server, which can return tool schemas/recommended plans as readable text that the agent ingests into its LLM context; this is outsider-provided content (public/third-party MCP output) rather than user-authored data.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill requires adding and calling the MCP endpoint at https://rube.app/mcp during runtime (via RUBE_SEARCH_TOOLS / RUBE_MANAGE_CONNECTIONS) to fetch tool schemas and execution plans, and those fetched artifacts directly control agent prompts and tool execution.