Content
62%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-organized, actionable workflow body with a clear phased sequence and error recovery, weakened by repeated quality-criteria prose and placeholder-based code examples. Progressive disclosure is structure-only with no file-level split.
Suggestions
De-duplicate the biological-replicate/metadata quality criteria that appears in both 'Domain Reasoning' and 'Reasoning Framework' into a single canonical statement.
Replace placeholder code (e.g. accession=accession, keywords="[gene/disease]") with concrete, runnable examples or clearly mark the values to substitute.
Move the dense 'Tool Reference' and 'Search Parameters' sections into a reference file referenced from SKILL.md to improve progressive disclosure.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly lean and table-driven, but quality criteria ('Prioritize >=3 biological replicates, complete annotations, both raw+processed data') and the RNA-seq/microarray comparison are restated across 'Domain Reasoning' and 'Reasoning Framework', offering tightening room. | 2 / 3 |
Actionability | Provides concrete tool-call signatures, fallback chains, and search parameters, but the code blocks use placeholders (keywords="[gene/disease]", accession=accession) making them templates rather than fully executable copy-paste examples. | 2 / 3 |
Workflow Clarity | Clear Phase 0→1→2→3 sequence with a diagram, an error-handling table supplying recovery feedback loops, and quality tiers plus synthesis questions acting as a checklist for a non-destructive retrieval task. | 3 / 3 |
Progressive Disclosure | Well-sectioned with clear headers, but ~145 lines are entirely inline with no bundle files or reference split; the dense Tool Reference and Search Parameters sections could be separated for cleaner disclosure. | 2 / 3 |
Total | 9 / 12 Passed |