hugging-face-vision-trainer
Pass
Audited by Gen Agent Trust Hub on Mar 12, 2026
Risk Level: SAFE
Full Analysis
- Authentication and Secret Management: The skill manages Hugging Face authentication tokens (HF_TOKEN) for model persistence. It correctly instructs users to pass these as job secrets, ensuring they are handled securely within the Hugging Face Jobs infrastructure.
- Dependency and Script Integrity: The training templates include PEP 723 inline metadata to specify their execution environment. The dependencies listed are well-known, trusted libraries from official registries (e.g., Transformers, Datasets, Timm, and PyTorch).
- Secure Data Handling: The dataset inspection tool uses official Hugging Face Datasets Server APIs to validate data structures. This helps ensure that training data is compatible and correctly formatted before compute resources are allocated.
- Standard Workflow Patterns: The skill implements standard machine learning workflows, including experiment tracking and Hub persistence, using official tools and documented practices with no evidence of unexpected or malicious behavior.
Audit Metadata