Content
71%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.
A well-organized, highly actionable skill body: method-by-method executable R code with interpretation thresholds and a consolidated sensitivity framework. Its weaknesses are the absence of any progressive disclosure (everything is inlined in one long SKILL.md with no reference files) and a few small correctness/organization blemishes such as the contamination-mixture/MR-RAPS label mix-up and the vignette-deferred multivariable section.
Suggestions
Split the long tail into one-level-deep reference files (e.g. references/sensitivity-framework.md for the run_sensitivity/summarize functions, references/strobe-mr-reporting.md for the checklist) so SKILL.md stays a lean overview.
Fix the mislabeled block in Additional Sensitivity Methods: the '# Contamination mixture' comment is followed by mr_raps code; either show the actual contamination-mixture call or correct the label.
Add a short decision guide for combining the tests (e.g. if Egger intercept is significant but PRESSO global is not, prefer Egger-adjusted estimates; if distortion test is significant, report corrected IVW) to make the workflow across methods explicit.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by executable R code with terse, decision-relevant comments; the pleiotropy-type taxonomy is domain knowledge that maps directly to the tests rather than padding. Minor trimmables (repeated MR-PRESSO setup in the framework function, the boilerplate Version Compatibility block) fit the 4 anchor; not 5 due to that redundancy, not 3 since no section extensively re-teaches known concepts. | 4 / 5 |
Actionability | Each method ships concrete, near copy-paste R code with named functions, parameters, and exact result-field access (e.g. 'presso_result$`MR-PRESSO results`$`Global Test`$Pvalue'). Minor gaps keep it at 4 rather than 5: the multivariable MR block defers to 'See TwoSampleMR vignette' without code, and the 'Contamination mixture' comment mislabels a code path that actually calls mr_raps. | 4 / 5 |
Workflow Clarity | Each section pairs a Goal/Approach header with embedded interpretation checkpoints (I^2 > 0.9 warning, p < 0.05 outlier cutoff, distortion-test meaning), and the comprehensive framework sequences eight analyses into one function with a summary step. It sits at 4 rather than 5 because there is no explicit ordering or decision guidance across methods (e.g. what to do when the Egger intercept is significant but the PRESSO global test is not). | 4 / 5 |
Progressive Disclosure | The ~300-line body is entirely inline with no reference files at all; the comprehensive sensitivity framework, STROBE-MR checklist, and multivariable MR section are candidates for separate reference documents. Clear section headers keep it above the 2 anchor's unstructured inline dump, but nothing is offloaded to navigable references, matching the 3 anchor. | 3 / 5 |
Total | 15 / 20 Passed |