GroqCloud Automation
Warn
Audited by Gen Agent Trust Hub on Feb 14, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- External Downloads (MEDIUM): The skill connects to a remote MCP server at
https://rube.app/mcp. This is an external dependency from an unverified source, creating a risk of unauthorized code execution or data exposure via the third-party bridge. - Prompt Injection (MEDIUM): The skill is vulnerable to indirect prompt injection due to its processing of untrusted external content.
- Ingestion points: Untrusted data enters the agent via the
messagesparameter inGROQCLOUD_GROQ_CREATE_CHAT_COMPLETIONand via thefile_pathparameter (which supports HTTP URLs) inGROQCLOUD_GROQ_CREATE_AUDIO_TRANSLATION. - Boundary markers: There are no delimiters or instructions provided to the agent to treat this data as untrusted or to ignore embedded instructions.
- Capability inventory: The skill generates AI responses used for agent reasoning and can perform network requests to fetch external audio files, potentially allowing for data exfiltration or SSRF if given a malicious URL.
- Sanitization: No input validation, escaping, or filtering is performed on the ingested content.
Audit Metadata