coinranking-automation

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION] (HIGH): Significant risk of Indirect Prompt Injection via external tool discovery. The skill instructs the agent to fetch 'recommended execution plans' and schemas from the external RUBE_SEARCH_TOOLS endpoint and use them to execute tools. This allows the external service to control the agent's logic flow and action sequence. Evidence Chain: Ingestion point: RUBE_SEARCH_TOOLS response in SKILL.md; Boundary markers: Absent; Capability inventory: RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH; Sanitization: Absent.
  • [EXTERNAL_DOWNLOADS] (MEDIUM): Requires connection to an untrusted external endpoint https://rube.app/mcp. This source is not listed in the trusted repositories or organizations, posing a supply-chain risk for the MCP server content.
  • [COMMAND_EXECUTION] (HIGH): The skill provides access to RUBE_REMOTE_WORKBENCH, which facilitates remote execution. When combined with dynamic plans from an untrusted source, this presents a high risk of unauthorized command execution.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 07:44 AM