The Agent Skills Directory

[SAFE]: The skill is a comprehensive reference for TorchDrug, a legitimate open-source library for drug discovery. The instructions, model architectures, and dataset descriptions align with standard scientific research practices and the library's official documentation. No malicious patterns or deceptive instructions were identified.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external chemical and biological data, including SMILES strings and PDB protein structures, via methods described in references/core_concepts.md and references/protein_modeling.md. While the skill does not explicitly define boundary markers to delimit these inputs, it utilizes robust, library-native parsers from RDKit and TorchDrug for data sanitization. The capabilities associated with this data are limited to model training and property prediction, with no evidence of dangerous operations like arbitrary file writing or shell command execution.\n- [EXTERNAL_DOWNLOADS]: The skill facilitates the installation of the torchdrug library and its dependencies from standard package registries. It also references official resources from the DeepGraphLearning GitHub organization and the torchdrug.ai domain, which are reputable sources for this scientific toolkit.

torchdrug