MiniMax
Pass
Audited by Gen Agent Trust Hub on Apr 20, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns, obfuscation, or unauthorized exfiltration were detected. The skill instructions emphasize operational safety, including budget boundaries and rights verification before data processing.
- [PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection due to its handling of untrusted user data and external documentation.
- Ingestion points: Processes user-provided text scripts (
speech-workflows.md), media references (media-generation.md), and official API documentation via external URL lookups (text-interfaces.md). - Boundary markers: Relies on procedural safeguards (user approval steps) and lacks automated technical delimiters for separating untrusted data from instructions.
- Capability inventory: File system write access in
~/minimax/for state persistence, network connectivity to official MiniMax API endpoints viacurl, and potential interaction with remote MCP servers. - Sanitization: Includes specific instructions to validate media rights, formats, and language settings before proceeding with generation tasks.
Audit Metadata