If you're doing single-cell analysis with deep learning, this is your toolkit. It covers the scvi-tools ecosystem: scVI for batch correction, scANVI for label transfer, totalVI for CITE-seq, MultiVI for multiome data, and DestVI for spatial deconvolution. The skill includes workflow-specific reference files and CLI scripts that handle the tedious bits like data prep and clustering. The main thing to know is that these models need raw counts, not normalized data, and you'll want to select highly variable genes first. The decision tree alone saves you from reading through model papers to figure out which one fits your data type.
npx skills add https://github.com/anthropics/knowledge-work-plugins --skill scvi-tools