The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the 'PyTDC' package from PyPI and downloads curated drug discovery datasets from the official Therapeutics Data Commons servers (tdcommons.ai) and associated GitHub repositories. These are well-known scientific resources and represent the intended primary purpose of the skill.
[INDIRECT_PROMPT_INJECTION]: The skill processes external biological and chemical data (SMILES strings, protein sequences) ingested from the TDC platform.
Ingestion points: Data is loaded via the 'tdc' library in scripts such as 'load_and_split_data.py' and 'molecular_generation.py'.
Boundary markers: Data is treated as structured numerical or string-based scientific entities rather than natural language instructions.
Capability inventory: Capabilities are limited to local data processing using 'pandas' and 'numpy', and writing datasets to a local 'data/' directory. No arbitrary shell execution or network exfiltration from the data content was detected.
Sanitization: Datasets are sourced from a curated scientific platform, though users should apply standard data validation when training models on external chemical space.
[COMMAND_EXECUTION]: The skill provides instructions for installing dependencies using the 'uv' package manager. These commands are standard for setting up the Python environment required for the TDC library.

pytdc