Content
63%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.
An exceptionally actionable skill body with concrete, copy-paste-ready commands and genuine domain gotchas (AEPT filtering trap, R-vs-Python spline divergence, per-gene aggregation), but it is noticeably verbose: the same rules are restated three to four times and generic statistics tutorial content is inlined. Several referenced bundle files do not exist, undermining navigation.
Suggestions
Deduplicate: state the AE/AEPT rule, spline guidance, and expression_anova usage once each and cross-reference, removing the repeated Analysis conventions and Bundled Scripts restatements.
Fix or remove the dead references — anova_and_tests.md, common_patterns_summary.md, TOOLS_REFERENCE.md, QUICK_START.md, and EXAMPLES.md are cited but absent from the bundle.
Move the generic statistics tutorial content (model-selection decision tree, Interpretation Guidance, Evidence Grading, statsmodels-vs-sklearn table) into a reference file; Claude already knows these concepts.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The 687-line body duplicates its own guidance repeatedly — the AE/AEPT no-filter rule appears in the PRIMARY SCRIPTS table, the CRITICAL section, and again in Analysis conventions; spline guidance appears four times; two identical expression_anova.py command blocks are pasted verbatim. It also inlines concepts Claude already knows ("OR = 1.0 means no association", "HR > 1 means increased hazard", the model-selection decision tree, the statsmodels-vs-scikit-learn table, Evidence Grading). This matches anchor 2 ('several unnecessary explanations or padded sections') rather than 3, where padding would be incidental rather than structural. | 2 / 5 |
Actionability | Guidance is fully executable and copy-paste ready: complete CLI invocations with all flags (e.g. the r_natural_spline_regression.py and logistic_regression_or.py commands), worked `tu run` examples with expected outputs, wrong-vs-right code snippets, and a stdlib fallback for missing scipy. Specific examples cover the common cases, matching the anchor for a 5. | 5 / 5 |
Workflow Clarity | There is a clear sequenced workflow — RULE ZERO (check pre-computed results first) → Phase 0 data validation → Phase 1 model fitting → Phase 2 diagnostics → Phase 3 interpretation — with explicit checkpoints: Phase 0 column/type checks, sanity heuristics that act as feedback loops ("If your F-statistic is >10 ... you probably aggregated"), and a Completeness Checklist. It falls short of 5 because the sequence is scattered across duplicated sections (PRIMARY SCRIPTS, CRITICAL, Analysis conventions) and precedence between sections must be synthesized by the reader. | 4 / 5 |
Progressive Disclosure | References are one level deep and clearly signaled (dedicated Support section), and all 6 references/*.md files exist — but the body cites anova_and_tests.md (4 times), common_patterns_summary.md, TOOLS_REFERENCE.md, QUICK_START.md, EXAMPLES.md, and test_skill.py, none of which exist in the bundle, so navigation dead-ends. Additionally ~200 lines of reference-grade tutorial content (decision tree, interpretation guidance, evidence grading) are inlined in SKILL.md. This fits anchor 3 ('some structure but could be better organized; content that should be separate is inline') — real structure exists, so not 2, but the broken paths exceed the 'minor organization gaps' of 4. | 3 / 5 |
Total | 14 / 20 Passed |