stable-diffusion-image-generation
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill follows standard practices for using well-maintained machine learning libraries from trusted organizations like Hugging Face, Meta (xformers), and StabilityAI.
- [SAFE]: External downloads and references (e.g., Hugging Face Hub, PyTorch registries) are directed toward well-known, established technology services.
- [INDIRECT_PROMPT_INJECTION]: The skill provides an interface for processing untrusted user input (prompts and images) through diffusion pipelines.
- Ingestion points: User-provided text prompts and initialization/mask images in
SKILL.mdandreferences/advanced-usage.md. - Boundary markers: Absent in the provided boilerplate code.
- Capability inventory: The skill facilitates file system writes (
image.save) and network communication for retrieving model weights and deploying APIs. - Sanitization: No specific prompt sanitization is implemented; the troubleshooting guide notes that
pipe.safety_checker = Nonecan be used to bypass model safety filters, which should be evaluated by the user based on their deployment context.
Audit Metadata