Content
86%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.
An excellent reference-style skill: fully executable code, clear tool-vs-code decision guidance, and a well-structured one-level-deep bundle split for specialized domains. The only weaknesses are minor duplication of pagination patterns between Sections B and D and the absence of an explicit validation checkpoint after bulk fetches.
Suggestions
Remove the per-API pagination loops in Section B (#2, #8, #9) or reduce them to one-line pointers to the generic pagination patterns in Section D to cut duplicated code.
Add an explicit post-batch validation checkpoint, e.g., after a bulk fetch assert len(all_records) matches the reported hitCount/total before parsing, since the skill targets batch operations with thousands of records.
Consider offloading the Section C restricted-sources table into the existing reference file to further slim the main SKILL.md, keeping only the most common restricted sources inline.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is a dense, prose-free cookbook that assumes Claude's competence — every line is a runnable pattern or a table row. Not 5: pagination appears both in Section B examples (#2 UniProt, #8 ClinicalTrials.gov, #9 EuropePMC) and again generically in Section D, a duplication that could be trimmed. | 4 / 5 |
Actionability | Fully executable copy-paste code throughout: real API endpoints (eutils, GDC, GTEx storage), working filter payloads, complete batch-fetch loops, and a runnable retry helper. Not 4: there are no pseudocode gaps; the examples cover the common cases end-to-end. | 5 / 5 |
Workflow Clarity | A decision table (Tool vs Code), explicit search->filter->download->parse sequences, and an error-handling section with status-code checks and an HTML-error-page guard provide most checkpoints. Not 5: batch/bulk workflows lack an explicit post-fetch validation step (e.g., verify expected record count before parsing). | 4 / 5 |
Progressive Disclosure | SKILL.md keeps core patterns inline and offloads 14 specialized domains to a real, one-level-deep reference (references/specialized-domains.md), clearly signaled via 'read references/specialized-domains.md when you need...' plus a When-to-Read summary table. Not 4: the split is appropriate and navigation is easy, with no buried or nested references. | 5 / 5 |
Total | 18 / 20 Passed |