pytorch
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill promotes the use of
torch.load(..., weights_only=True), which is a critical security best practice in the PyTorch ecosystem to prevent arbitrary code execution via malicious pickle files. - [SAFE]: Instructions explicitly warn against loading untrusted full pickles and emphasize using
state_dictfor checkpoints, further reducing the risk of malicious payload execution. - [SAFE]: The provided code snippets follow standard development patterns for device placement, mixed precision (AMP), and evaluation modes without any hidden or suspicious operations.
- [SAFE]: No evidence of prompt injection, data exfiltration, obfuscation, or unauthorized network access was detected in the skill's instructions or metadata.
Audit Metadata