Content
62%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 skill's core strength is an exemplary seven-phase workflow with genuine validation loops, error recovery, and a quality checklist. It is undermined by token inefficiency — large sections duplicate the bundled reference files and the summary/integration sections pad the document — and by progressive-disclosure failures including a reference to a nonexistent generate_schematic.py script.
Suggestions
Replace the inline "Database-Specific Search Guidance" and "Citation Style Guide" sections with one-line pointers to references/database_strategies.md and references/citation_styles.md, keeping only a single quick-start example inline.
Remove the mandatory schematics section's reference to scripts/generate_schematic.py (which is not bundled) or bundle the script; also delete the redundant closing Summary and tangential Integration sections to cut token load.
Provide a concrete arXiv/Semantic Scholar API example (e.g., an actual curl or Python request) so the "export results in JSON" step is executable rather than aspirational.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is noticeably verbose with several padded/duplicated sections: "Database-Specific Search Guidance" (~70 lines) and "Citation Style Guide" restate content already in references/database_strategies.md and references/citation_styles.md; the closing "Summary" section restates the whole skill; "Integration with Other Skills" lists tangential skills (pydeseq2, scanpy, brand-guidelines); the schematics section repeats its own instructions ("How to generate figures" vs "How to generate schematics"); and the same script invocations (search_databases.py, verify_citations.py) are shown three times across Phase sections and the Example Workflow. This fits anchor 2 ("noticeably verbose; several unnecessary explanations or padded sections") better than 3, given the volume of duplication of bundled reference content. | 2 / 5 |
Actionability | Mostly executable guidance: concrete gget commands, copy-paste invocations with flags for the three bundled scripts (e.g., `python scripts/verify_citations.py my_literature_review.md`, `--deduplicate --format markdown`), a template cp command, and dependency checks. Minor gaps keep it below anchor 5: `scripts/generate_schematic.py` is invoked but does not exist in the bundle, the arXiv section shows only commented pseudocode with no actual API call, and "Export results in JSON format from each database" gives no concrete mechanism. | 4 / 5 |
Workflow Clarity | The seven-phase workflow is clearly sequenced with explicit validation checkpoints and feedback loops: "Re-run verification until all citations pass", a citation verification script, database-failure handling ("If fewer than 2 databases succeed, pause and inform user"), a dedup-failure fallback ("Manually deduplicate by sorting on title and DOI"), `--check-deps` for PDF tooling, and a final quality checklist. This matches anchor 5 (explicit validation steps, error-recovery loops, checklist). | 5 / 5 |
Progressive Disclosure | Structure exists and the bundled references are real, one level deep, and listed in a Resources section, but substantial content that belongs in those files is inlined: the Database-Specific Search Guidance and Citation Style Guide sections duplicate references/database_strategies.md and references/citation_styles.md, and the body points to a nonexistent scripts/generate_schematic.py. This fits anchor 3 ("content that should be separate is inline") better than 4, given the scale of the inline duplication. | 3 / 5 |
Total | 14 / 20 Passed |