Content
53%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.
A thorough, well-referenced skill with strong query examples and executable API code, undermined by significant duplication between SKILL.md and its own reference files plus across its internal sections. The workflows are clearly sequenced but lack validation checkpoints for the batch operations they perform.
Suggestions
Cut duplicated content: state rate limits, search-history expiry, and clipboard limits once and remove the redundant repetitions in the Tips, API Usage, and Limitations sections; trim sections 7–8 (My NCBI, RSS, clipboard) to one-line pointers.
Move the MeSH subheading list, publication-type list, and detailed API parameter documentation into the existing reference files, keeping SKILL.md as an overview with example queries and the two core code blocks.
Add validation checkpoints to Workflow 3 (Programmatic Data Extraction): check HTTP response status and esearchresult count before fetching, implement the exponential-backoff retry on HTTP 429 that the tips recommend, and verify record count after EFetch.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The ~460-line body is noticeably verbose: rate limits appear three times (§4, Tips, Limitations), search-history expiry and clipboard limits each appear twice, and sections 7–8 explain web-UI features (My NCBI, RSS feeds, clipboard) tangential to what the skill is invoked for. The 'mostly efficient' bar of 3 is exceeded by this padding, though the core query/API detail is genuinely non-obvious domain knowledge. | 2 / 5 |
Actionability | Two complete, executable Python blocks cover the common search-then-fetch cases, and example queries are copy-paste ready. Minor gaps: the elink and batch-export examples are bare URL query strings rather than runnable requests, and exponential backoff is recommended but never shown. | 4 / 5 |
Workflow Clarity | Five numbered workflows are clearly sequenced, but the programmatic workflows involve batch operations (EPost, bulk EFetch, batch export) with no validation or verification checkpoints — no response-status checks, retry-on-failure, or verify-output steps. Per the batch-operations rule, workflow clarity is capped at 3. | 3 / 5 |
Progressive Disclosure | Reference signaling is excellent: per-section 'When to consult X.md' bullets, grep patterns for locating content, and a dedicated 'Working with Reference Files' section with a loading strategy; references are one level deep and real. The gap is that the body inlines substantial detail (MeSH subheading lists, publication types, API parameters) that duplicates the reference files, keeping it below the 'appropriately split' bar of 5. | 4 / 5 |
Total | 13 / 20 Passed |