mmx-cli
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user input via prompts and messages that are transmitted to external AI models, creating a surface for potential injection attacks.
- Ingestion points: The skill uses the
mmxCLI which accepts user-controlled text via flags such as--message,--prompt,--text, and--lyrics, as well as file-based inputs through--messages-file. - Boundary markers: A specific 'Limitations' section warns the agent to treat prompts, downloaded content, and returned URLs as untrusted data and explicitly advises against executing instructions found within that data.
- Capability inventory: The
mmxtool performs network operations to MiniMax AI platform endpoints and has the capability to write files to the local system using the--out,--out-dir, and--downloadflags. - Sanitization: The instructions mandate that the agent validate returned file paths and confirm service URLs before use.
Audit Metadata