nocrm-io-automation
Fail
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [Unverifiable Dependencies & Remote Code Execution] (HIGH): The skill directs users to add an external MCP server 'https://rube.app/mcp'. This server is not a trusted source and dynamically provides the tools and execution logic for the agent.
- [Indirect Prompt Injection] (HIGH): The skill processes untrusted data from Nocrm IO (CRM leads/contacts) and has the capability to perform write operations via 'RUBE_MULTI_EXECUTE_TOOL'. Evidence: 1. Ingestion points: Data returned from Nocrm IO toolkit tools. 2. Boundary markers: Absent. 3. Capability inventory: 'RUBE_MULTI_EXECUTE_TOOL' and 'RUBE_REMOTE_WORKBENCH' enable file/data modification. 4. Sanitization: No sanitization or validation of CRM content is mentioned.
- [Data Exposure & Exfiltration] (MEDIUM): Interaction with sensitive CRM data occurs through 'rube.app', a non-whitelisted domain. While the protocol is MCP, the destination is an untrusted external endpoint, posing a risk of data transit through a third-party proxy.
Recommendations
- AI detected serious security threats
Audit Metadata