This handles the statistical heavy lifting of GWAS fine-mapping: taking a lead SNP and figuring out which variant in the LD block is actually causal, not just the best-tagged proxy on the genotyping array. It wraps Open Targets Genetics and GWAS Catalog APIs to pull credible sets from SuSiE and FINEMAP, then connects variants to genes using locus-to-gene scores that factor in eQTL colocalization and chromatin data. The reasoning framework is the real value here: it walks you through LD structure interpretation, posterior probability thresholds, and why the nearest gene is usually wrong. If you're doing target identification from GWAS hits and tired of people assuming the lead SNP is causal, this codifies the right questions to ask.
npx skills add https://github.com/mims-harvard/tooluniverse --skill tooluniverse-gwas-finemapping