Content
63%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The body is a well-structured, highly actionable API guide — concrete tool parameters, executable JSON examples, and a sensible phased workflow. Its weaknesses are efficiency and organization: a large interpretation section restates textbook population genetics, and bulky reference tables are inlined instead of split into bundle files.
Suggestions
Move the population-code table, data-collections table, and the Reasoning Framework (evidence grading, Fst/LD interpretation guidance) into a references/ file (e.g. references/population-codes.md and references/interpretation.md), keeping only a pointer and the synthesis questions in SKILL.md.
Trim the interpretation guidance to only the non-obvious specifics (e.g. 1000 Genomes sample-size precision limits, when to fall back to gnomAD) and cut textbook definitions of Fst, LD, and genetic diversity that Claude already knows.
Make workflow Step 4 concrete: show the actual cross-reference step (e.g. a pandas merge of GWAS hits with per-population frequencies) or a short Python analysis template, which would also model the COMPUTE, DON'T DESCRIBE mandate.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Tool documentation is tight and operational, but the "Reasoning Framework for Result Interpretation" section re-teaches textbook genetics Claude already knows ("Fst measures population differentiation (0 = no differentiation, 1 = complete fixation)", "AFR populations have the highest genetic diversity", "LD patterns differ by ancestry"). Mostly efficient with a substantial block of unnecessary explanation — the 3 anchor, not 2 (padding is confined to one section) and not 4 (that block is well over minor). | 3 / 5 |
Actionability | Every tool gets concrete parameter specs, response schemas, and copy-paste JSON payloads (e.g. {"superpopulation": "AFR", "limit": 10}), which is mostly executable guidance. However workflow Step 4 ("Cross-reference with population data for stratification analysis") is direction without steps, and the "COMPUTE, DON'T DESCRIBE" mandate demands Python analysis but shows no analysis code example — minor gaps, matching the 4 anchor rather than the 5 anchor's fully covered common cases. | 4 / 5 |
Workflow Clarity | The population-stratification workflow gives a clear 4-step sequence with a concrete tool call for each step, and the "NOT for" section routes to sibling skills. Checkpoints are only implicit and Step 4 is underspecified, so it matches the 4 anchor (clear sequence, minor gaps) rather than 5 (no explicit validation or error-recovery loop); operations are read-only, so the destructive/batch cap at 3 does not apply. | 4 / 5 |
Progressive Disclosure | No bundle files exist (no references/, scripts/, or assets/), and all reference material is inlined in a ~220-line body: the 14-row population-code table, the data-collections table, and the ~40-line interpretation framework each belong in a one-level-deep reference file. Section headers are clear and navigation within the file is fine (so not the 2 anchor's minimal structure), but content that should be separate is inline — the 3 anchor. | 3 / 5 |
Total | 14 / 20 Passed |