The Agent Skills Directory

[SAFE]: The skill and its reference files act as an educational resource for the legitimate TorchDrug library. The provided code snippets follow standard practices for machine learning development, including dataset loading, model architecture definition, and training loops.
[EXTERNAL_DOWNLOADS]: The skill instructs users to install the torchdrug package using standard package managers like uv. This is a standard and expected operation for the skill's purpose. The library is a well-known open-source project in the chemical machine learning community.
[COMMAND_EXECUTION]: Code examples demonstrate the use of standard Python and PyTorch APIs. There are no instances of unsafe shell command execution, dynamic code evaluation of untrusted strings, or privilege escalation attempts.
[DATA_EXFILTRATION]: While the skill mentions processing biological and chemical data, all data handling is performed locally or through standard library-managed dataset downloads. No network exfiltration or unauthorized sensitive data access patterns were identified.

torchdrug