The body is mostly efficient, but the Overview and 'Key Concepts' sections re-explain GWAS, linkage disequilibrium, credible sets, posterior probability, and L2G — concepts a capable model already knows — which is padding the rubric penalizes.

Concrete tool names and PP/L2G thresholds are given, but despite the 'COMPUTE, DON'T DESCRIBE' mandate there is no executable code anywhere — tool calls appear only as prose names with no example invocations or arguments.

The 5-step reasoning framework is clearly sequenced with conditional logic at each step, but there are no explicit validation checkpoints, feedback loops, or checklists — the decision points are implicit rather than enforced.

The ~170-line body is monolithic with no bundle/reference files; content that could be split out (the tool reference list, FAQ, and output-interpretation bands) is inline rather than one level deep in signaled reference files.

Lists multiple concrete actions — 'Statistical fine-mapping of GWAS loci using credible sets (SuSiE, FINEMAP)', 'locus-to-gene scoring (Open Targets L2G)', and 'prioritizing causal variants…comparing fine-mapping methods…converting lead SNPs to target genes'.

Explicitly answers both what it does and when to use it via the 'Use for prioritizing causal variants at GWAS hits, comparing fine-mapping methods, and converting lead SNPs to target genes' clause.

Uses natural domain terms a researcher would actually say — 'fine-mapping', 'GWAS', 'causal variants', 'lead SNPs', 'target genes' — with good coverage and no over-reliance on jargon.

Carves a clear niche (credible-set fine-mapping + L2G) and explicitly distinguishes itself from positional 'nearest gene' approaches, with triggers unlikely to fire for adjacent skills.