Content
68%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 highly actionable, example-driven reference with real, existing bundle files and clear pointer navigation, but it is overlong for a SKILL.md: best-practice material duplicates the capability examples, and API-reference content is inlined rather than kept in references/api_guide.md. Batch and large-extraction workflows also lack validation checkpoints, capping workflow clarity at 3.
Suggestions
Consolidate the 'Critical Best Practices' section: replace the repeated code blocks with one-line cross-references to the capability examples, and delete the unexplained 'LitLLM with OpenRouter / don't use Perplexity API' note from the closing Notes.
Move the 'Common Filter Patterns', 'Entity Types', and 'External IDs' sections into references/api_guide.md, leaving only a one-line pointer so SKILL.md stays a lean overview.
Add validation steps to the batch and large-extraction workflows: verify returned result counts against requested totals (batch_lookup, paginate_all max_results), check for partial failures, and instruct retry/backoff behavior on incomplete data before exporting to CSV.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense with executable code and avoids explaining concepts Claude already knows, but the 'Critical Best Practices' section re-explains patterns already demonstrated in the capabilities section (two-step lookup, per-page size, batching, sampling), and the unexplained closing note about 'Use LitLLM with OpenRouter... don't use Perplexity API directly' is padding. Mostly efficient but could be tightened. | 3 / 5 |
Actionability | Every capability ships copy-paste-ready, executable code with realistic arguments (e.g., client.search_works with filter_params and sort, find_author_works(author_name="Jennifer Doudna"), batch lookups with real DOIs, pagination, and a complete CSV export loop), covering the common cases comprehensively. | 5 / 5 |
Workflow Clarity | The two-step entity-lookup pattern is explicitly sequenced and Troubleshooting provides error-to-fix feedback (403 → verify email/rate; empty results → check filter syntax), but the batch operations (batch_lookup of up to 50 IDs, paginate_all with max_results=10000, 25k samples) have no validation or verification steps such as checking returned counts or handling partial failures. The batch-operations cap holds this at 3. | 3 / 5 |
Progressive Disclosure | Good structure: overview and capabilities in the body with well-signaled, one-level-deep references ('See references/api_guide.md for: Complete filter syntax, All available endpoints...'; 'See references/common_queries.md for...'), both of which exist. However, the inline 'Common Filter Patterns', 'Entity Types', and 'External IDs' sections duplicate api_guide.md material that belongs in the reference file, so the split is not fully clean. | 4 / 5 |
Total | 15 / 20 Passed |