wtf is ml-shap-skill?

ML-SHAP Skill is a reusable add-on for Codex, an AI coding assistant, that runs a complete machine learning analysis on tabular data stored in CSV or Excel files. Instead of writing all the analysis code yourself, you point the skill at your data and describe your goal, and it handles the statistical workflow for you. The skill first checks the data for common problems: missing values, duplicate rows, inconsistent types, outliers, and signs that information from outside the training set has leaked into it. It then picks an appropriate way to split the data into training and test portions depending on whether the data is a plain snapshot, follows a timeline, is grouped by category such as city or customer, or has a spatial structure. From there it compares several types of models, from simple baselines to tree-based and boosting models, and tunes them carefully so the test portion is never used to influence the tuning. A major feature is explainability through SHAP, a method for showing how much each input variable contributed to a prediction and in which direction. The skill produces SHAP charts such as beeswarm plots showing feature importance and value spread, along with dependence and interaction charts. When SHAP cannot be used for a particular model, it falls back to alternative importance methods and clearly labels them as substitutes rather than presenting them as SHAP results. The output is meant to be publication-ready: charts saved as PNG, PDF, and SVG, an Excel workbook, a written report, and the underlying reproducible code, plus an optional JSON summary. The skill is careful about scientific correctness too, for example refusing to describe same-period estimates as future forecasts and avoiding causal claims from what are simple predictive explanations. To install it, you copy the ml-shap folder into your local Codex skills directory and restart Codex so it becomes available to invoke by name. No license has been specified yet, so the author asks that one be added before the skill is redistributed or reused elsewhere.