minimax-h3-prompting
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests user briefs and video metadata to generate structured prompts. It mitigates injection risks by using a strict, model-facing serialization protocol defined in
official-prompt-protocol.md. It utilizes specific delimiters like<d>for dialogue and[Shot N]for timing, which helps isolate user-provided content from the agent's instructional logic. - [EXTERNAL_DOWNLOADS]: The skill references technical documentation and capability reports from
github.com/MiniMax-AI,platform.minimaxi.com, andminimax.io. These resources are official and well-known sources for the model being prompted. The references are used to maintain consistency with the model's official API and prompt syntax.
Audit Metadata