If you're analyzing single-cell RNA-seq data, this gives Claude the full scanpy toolkit for the standard workflow: loading h5ad or 10X files, running QC on cell and gene counts, normalizing with log transforms, computing PCA and UMAP embeddings, running Leiden clustering at different resolutions, and identifying marker genes with Wilcoxon tests. The skill includes practical defaults like filtering cells with over 5% mitochondrial counts and using 2000 highly variable genes. It's built around the AnnData object structure, so Claude understands how to access expression matrices, metadata, and embeddings. Really solid for going from raw counts to annotated cell types without writing the boilerplate yourself.
npx skills add https://github.com/davila7/claude-code-templates --skill scanpy