The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill includes instructions to install the TorchDrug library and its dependencies (torch, torch-scatter, torch-cluster) using official pip and conda commands. It references pytorch-geometric.com, an established and well-known domain for hosting graph neural network binaries.
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it is designed to process external scientific data formats like SMILES strings, PDB files, and biomedical knowledge graph triples. 1. Ingestion points: Data is loaded via methods such as data.Molecule.from_smiles, data.Protein.from_pdb, and built-in dataset loaders like datasets.BBBP. 2. Boundary markers: The documentation does not specify the use of delimiters or 'ignore' instructions for the biological data processed. 3. Capability inventory: The skill utilizes the torchdrug library's capabilities for model training, file I/O for saving/loading representations, and network operations to download curated datasets. 4. Sanitization: The reference material focus is on scientific modeling rather than input sanitization for potential adversarial instructions embedded in data strings.

torchdrug