z-image
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill consists of legitimate technical documentation for the Z-Image (造相) model family, including specific guidance for the Turbo and Base variants. It provides verified facts sourced from official vendor documentation on GitHub and Hugging Face.
- [SAFE]: All external URLs point to trusted or well-known repositories, documentation sites, and scholarly archives (github.com, huggingface.co, arxiv.org, and tongyi-mai.github.io). These references are documented neutrally as sources of truth for the model's capabilities.
- [SAFE]: The instructions do not contain any prompt injection patterns, data exfiltration commands, or obfuscated content. The use of non-Latin characters (Chinese) is contextually appropriate for a model developed by Alibaba and trained on bilingual datasets.
- [SAFE]: Recommendations for running the model via tools like
diffusersandmfluxare standard industry practices for local inference and do not involve suspicious installation or execution patterns.
Audit Metadata