Explore skills

PyTorch Geometric (PyG) is for graph neural networks, supporting node/link/graph classification, message passing (GCN, GAT, GraphSAGE, GIN), heterogeneous graphs, neighbor sampling, and custom datasets. Use it specifically with torch_geometric, not for general NetworkX analytics or non-graph PyTorch models.

PyTorch-native graph neural networks for molecules and proteins, suitable for custom GNN architectures in drug discovery, protein modeling, or knowledge graph reasoning. It's best for custom model development, protein property prediction, and retrosynthesis; for pre-trained models and diverse featurizers, use deepchem, and for benchmark datasets, use pytdc.

UMAP is a fast nonlinear dimensionality reduction technique for high-dimensional data, useful for 2D/3D visualization, clustering preprocessing (HDBSCAN), and supervised/parametric learning.

Query the U.S. Treasury Fiscal Data REST API for federal financial data, no API key required. Access national debt, Treasury statements, securities auctions, interest and foreign exchange rates, savings bonds, or U.S. government revenue and spending statistics.

This skill processes and analyzes large tabular datasets (billions of rows) that exceed available RAM, excelling at out-of-core DataFrame operations, fast aggregations, and efficient visualization. It's ideal for working with massive CSV/HDF5/Arrow/Parquet files and performing quick statistics on big data.

Molecular featurization for ML with over 100 featurizers, including ECFP, MACCS, descriptors, and pretrained models (ChemBERTa). Converts SMILES to features for QSAR and molecular ML.

A Pythonic wrapper for RDKit, offering a simplified interface and sensible defaults. It's ideal for standard drug discovery tasks such as SMILES parsing, standardization, descriptors, fingerprints, clustering, 3D conformers, and parallel processing, returning native rdkit.Chem.Mol objects.

This skill is for time series machine learning tasks including classification, regression, clustering, forecasting, anomaly detection, segmentation, and similarity search. Use it when working with temporal data, sequential patterns, or time-indexed observations requiring specialized algorithms beyond standard ML approaches, particularly for univariate and multivariate time series.

Hugging Face Transformers for loading Hub models, running pipeline inference, text generation, and Trainer fine-tuning on NLP, vision, audio, and multimodal tasks. Use when working with AutoModel, pipelines, tokenizers, or TrainingArguments—not for general ML outside the Transformers library.

Microscopy data management platform for high-content screening and microscopy workflows. It allows accessing images via Python, retrieving datasets, analyzing pixels, managing ROIs/annotations, and batch processing.

Diffusion-based molecular docking predicts protein-ligand binding poses from PDB/SMILES, providing confidence scores and enabling virtual screening for structure-based drug design. It is not intended for affinity prediction.

A comprehensive skill for writing markdown and Mermaid diagrams. It establishes text-based diagrams as the default documentation standard, offering full style guides, 24 diagram type references, and 9 document templates for scientific documents, reports, analyses, or visualizations.