Content
67%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.
Highly actionable content with strong pitfall guidance and a clear workflow, undermined by missing reference files that break progressive disclosure and some repetition across sections.
Suggestions
Create the referenced files (ANALYSIS_PROCEDURES.md, CODE_REFERENCE.md, TOOLS_REFERENCE.md, QUICK_START.md) under references/, or remove the dangling references from the body.
Consolidate the three repeated rows-vs-unique-sites explanations into a single canonical section to reduce token cost.
Trim generic sections like 'Key Principles' and 'LOOK UP, DON'T GUESS' that restate what Claude already knows.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and rich in non-obvious pitfalls, but the rows-vs-sites distinction is restated three times and 'Key Principles'/'LOOK UP, DON'T GUESS' add some generic padding that could be trimmed. | 3 / 5 |
Actionability | Copy-paste-ready script invocation with concrete args, question-to-field mapping tables, and specific tool parameter guidance (e.g. GTEx uses gene_symbol NOT Ensembl ID, ensembl_get_regulatory_features needs NO chr prefix). | 5 / 5 |
Workflow Clarity | RULE ZERO checkpoint and an explicit sanity-check feedback loop ('Re-run with the other axis and compare') anchor a clear phased workflow; minor validation gaps in the later batch phases keep it just below 5. | 4 / 5 |
Progressive Disclosure | The bundled script reference is real, but the body references four markdown files (ANALYSIS_PROCEDURES.md, CODE_REFERENCE.md, TOOLS_REFERENCE.md, QUICK_START.md) that do not exist in the references/ bundle, leaving navigation broken. | 2 / 5 |
Total | 14 / 20 Passed |