hypeauditor-automation

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION] (HIGH): Vulnerable to Indirect Prompt Injection through the dynamic discovery mechanism. \n
  • Ingestion points: RUBE_SEARCH_TOOLS fetches external tool schemas and 'recommended execution plans' from rube.app (SKILL.md). \n
  • Boundary markers: Absent; the skill explicitly instructs the agent to follow the instructions and plans provided in the search response. \n
  • Capability inventory: RUBE_MULTI_EXECUTE_TOOL and RUBE_REMOTE_WORKBENCH allow execution of complex tasks based on this remote input. \n
  • Sanitization: Absent.\n- [EXTERNAL_DOWNLOADS] (MEDIUM): The skill directs users to configure a third-party MCP endpoint (https://rube.app/mcp) that is not part of the established trusted source list.\n- [REMOTE_CODE_EXECUTION] (MEDIUM): Tool logic and execution parameters are defined by the remote server at runtime. This 'discover-then-execute' pattern allows the remote server to effectively control the agent's functional capabilities and actions.
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 12:47 PM