minimax-h3-text-to-video
Pass
Audited by Gen Agent Trust Hub on Aug 10, 2026
Risk Level: SAFE
Full Analysis
- [PROMPT_INJECTION]: The skill instructions define a structured workflow for generating creative prompts. No patterns of safety filter bypass or behavior override were detected.
- [DATA_EXFILTRATION]: There are no network requests, hardcoded credentials, or access to sensitive system files. The skill explicitly instructs the agent to return text prompts only and not to submit API payloads or generation jobs.
- [REMOTE_CODE_EXECUTION]: The skill does not download or execute any external scripts or packages. It relies entirely on markdown instructions and templates.
- [COMMAND_EXECUTION]: No shell commands or system-level operations are present in the instructions or metadata.
- [INDIRECT_PROMPT_INJECTION]: The skill serves as a prompt engineering framework. While it processes user concepts, it lacks the necessary capabilities (such as network or file system access) to be leveraged for malicious outcomes through indirect injection.
Audit Metadata