curated-automation
Warn
Audited by Snyk on Feb 16, 2026
Risk Level: MEDIUM
Full Analysis
MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).
- Potentially malicious external URL detected (high risk: 0.80). The skill requires configuring and calling the Rube MCP endpoint at https://rube.app/mcp during runtime (via RUBE_SEARCH_TOOLS / RUBE_MULTI_EXECUTE_TOOL), which returns tool schemas and drives execution of remote tools—meaning external content can directly control agent instructions and execute code.
Audit Metadata