If you're building graph neural networks with PyTorch, this gets you up and running with PyTorch Geometric fast. It covers the core use cases: node and graph classification, link prediction, molecular property prediction for drug discovery, and handling 3D geometric data like point clouds. The skill includes patterns for mini-batch processing and multi-GPU training, which matters when you're working with large graphs. With 27.7K stars on GitHub and 285 installs, it's a solid resource for anyone doing deep learning on irregular structures. Most useful when you need to move beyond standard tensor operations into graph-structured data, whether that's citation networks, social graphs, or chemical structures.
npx skills add https://github.com/davila7/claude-code-templates --skill torch-geometric