The Agent Skills Directory

[SAFE]: The skill documentation correctly identifies and utilizes the official TorchDrug 0.2.1 API and associated scientific workflows.
[EXTERNAL_DOWNLOADS]: Installation instructions reference the official PyTorch Geometric wheel index (data.pyg.org) and the TorchDrug documentation site (torchdrug.ai). These are well-known and trusted services in the research community.
[COMMAND_EXECUTION]: Shell commands are provided for standard environment management tasks such as virtual environment creation and package installation using the 'uv' package manager.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process external molecular and protein data. It mitigates potential risks by explicitly instructing the agent to sanitize inputs and validate outputs, which are standard security practices for data-driven scientific agents.
[DYNAMIC_EXECUTION]: Documents the use of standard library features for loading model checkpoints and configurations (torch.load, json.load), which is essential functionality for machine learning workflows.

torchdrug