@wan-huiyanchecks…add-ml-feature-evaluator → master1 files · +2 −0updated 2mo agoAdds two companion ML skills to the Data & Analysis section:
ML Feature Evaluator — Structured 10-step go/no-go diagnostic for adding features to production ML models. Covers outcome gradient, coverage gaps, entropy/gain ratio, conditional MI, incremental CV AUC, 9-point temporal safety, and implementation planning with mandatory review.
ML Training Window Assessor — Assess whether an ML training window can be extended by adding a new data source. Covers per-output label validity, drift-aware validation (PSI), purged temporal CV with embargo, XGBoost NaN handling for missing feature blocks, and companion model vs extended training architecture decisions.
Both skills are research-grounded (thresholds cite Quinlan, Brown et al., DeLong, Siddiqi, van der Ploeg, Chen & Guestrin) and battle-tested on production pipelines.
🤖 Generated with Claude Code
Thanks for the suggestion! @accesspath27-lgtm I've shortened both descriptions as suggested. Can you please take another look?