This connects GWAS hits to actual drug targets by doing the full chain: variant to gene (via fine-mapping and eQTL), gene to druggable protein (via OpenTargets tractability), protein to existing compounds (via ChEMBL and DGIdb). The workflow is smart about the causal direction problem: a loss-of-function variant protecting against disease means you want an inhibitor, not an agonist. It includes a composite scoring system weighing genetic evidence, druggability, clinical data, and novelty. The documentation is unusually good about parameter gotchas (ensemblId vs ensemblID, disease_trait not trait). Best for target validation when you have genetic evidence and want to skip straight to what's druggable and whether something already exists for repurposing.
npx skills add https://github.com/mims-harvard/tooluniverse --skill tooluniverse-gwas-drug-discovery