Content
90%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: dense, executable code with real endpoints and non-obvious operational knowledge (pagination cursors, rate-limit retries, HTML error-page guards), a clear tool-vs-code decision table, and a well-signaled one-level-deep reference file for specialized domains. The main improvement opportunities are a marginally long main body and the absence of one fully worked, checkpoint-validated end-to-end workflow.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is lean code-first reference material: every block carries non-obvious domain knowledge ("# VCF (no cyvcf2 needed)", "# Guard against HTML error pages (CDC, NCBI return 200 with HTML for missing files)", batch sizes, cursor-pagination details) with no explanations of concepts Claude already knows. It matches the 'every token earns its place' anchor; there is no padded prose to trim for score 4's 'minor instances'. | 5 / 5 |
Actionability | All guidance is fully executable, copy-paste-ready Python with real endpoints and parameters: the NCBI esearch→efetch batching loop, UniProt cursor pagination via the Link header, the GDC filters JSON POST, the GTEx gct.gz URL, and the retry/backoff helper. It matches the anchor for executable code covering common cases; it is not score 4 because no example is pseudocode or missing key details. | 5 / 5 |
Workflow Clarity | Sequencing is present and validated — the "Decision: Tool vs Code" table, the named "search -> filter -> download -> parse" workflow, an end-to-end exemplar (NCBI esearch → idlist → batched efetch), and Section D error handling (status-code checks, HTML error-page guard, JSON decode) provide checkpoints, so the batch-operation cap does not apply. It falls short of anchor 5 because there is no single explicit numbered workflow with validate-and-recover checkpoints at each stage; it sits noticeably above anchor 3 since validation patterns are concretely shown. | 4 / 5 |
Progressive Disclosure | The 14 specialized domains are correctly split into references/specialized-domains.md (verified to exist), referenced one level deep and clearly signaled by a table with a "When to Read" column. It falls short of anchor 5 because the main body is ~390 lines and sections like the 14-row restricted-sources table and the full format cookbook could arguably be split out as well — good structure with minor organization gaps, matching anchor 4 rather than a pure overview. | 4 / 5 |
Total | 18 / 20 Passed |