minimax-h3
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes local Python scripts for core logic, specifically
scripts/minimax_h3_skill_router.pyfor style classification andscripts/minimax_h3_prompt_validator.pyfor syntax validation. These are internal tools within the vendor's environment. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied "shot briefs" and external reference files (
base-en.txt,ref-en.txt). While this presents a potential ingestion surface for indirect prompt injection, the risk is mitigated by the "Deterministic format gate" which strictly validates H3 syntax and duration before execution. - [DATA_EXPOSURE]: The skill accesses project-specific directories for reading configuration and writing outputs (e.g.,
04_generation/prompts/minimax-h3/). All file access is restricted to relative paths within the expected project structure.
Audit Metadata