Content
92%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A highly actionable, well-sequenced epidemiology workflow with executable code, validation steps, and a checklist; its only weakness is that all material lives in a single SKILL.md with no progressive disclosure into bundle files.
Suggestions
Move the reusable code templates (download_and_parse, sample_size_logistic, the sequential-adjustment regression block) into scripts/ files and reference them one level deep, keeping SKILL.md as an overview.
Extract the detailed model-selection-by-outcome and assumption-check guidance into a references/ file (e.g. STATISTICAL_METHODS.md) linked from Step 5, reducing inline density.
Add a short 'References' section near the top signaling the bundle files so navigation is explicit rather than implicit.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and directive — almost every section leads with executable Python and brief, non-padded context ('Never describe what you "would do" — do it'), avoiding re-explaining concepts Claude already knows. | 3 / 3 |
Actionability | It provides copy-paste-ready, executable code throughout (data loading, regression, stratification, VIF, power analysis), with placeholder variables explicitly justified as adaptable ('adapt to your variables'), not pseudocode. | 3 / 3 |
Workflow Clarity | Nine steps are clearly sequenced with validation checkpoints (power analysis before dataset commitment, VIF/assumption checks, try/except in stratified runs) and capped by an explicit completeness checklist. | 3 / 3 |
Progressive Disclosure | Content is well-organized into sections but entirely monolithic — no bundle files exist and nothing is split into references/scripts, so ~260 lines of templates and code live inline where separation would aid discovery. | 2 / 3 |
Total | 11 / 12 Passed |