The Agent Skills Directory

[SAFE]: The skill provides instructional content and code snippets for PyTorch specialists. No prompt injection or behavior-overriding patterns were found in the routing logic or the markdown body.
[SAFE]: All code snippets follow standard development practices. The skill explicitly promotes security-conscious coding by warning about the risks of using torch.load with untrusted files and recommending the weights_only=True parameter.
[SAFE]: All library references (such as torch, numpy, boto3, and kornia) are standard, widely-used machine learning and cloud tools. There are no suspicious dependencies or calls to untrusted external domains.
[SAFE]: Command-line examples (e.g., torchrun, nvidia-smi, ping) are used appropriately for training orchestration and infrastructure debugging within the context of model development.
[SAFE]: No instances of obfuscation, sensitive data exposure, or persistence mechanisms were detected across the nine analyzed files.

using-pytorch-engineering