Initial release: splitguard v0.1.0 - runtime ML data-leakage detector Zero-config runtime detection of train-test leakage (overlap, preprocessing-fit, and the unique group-leakage case). Wraps scikit-learn / XGBoost / LightGBM / CatBoost fits, train_test_split and cross-validators. Honest about its limits; validated against the Yang et al. (ASE'22) benchmark. Full test suite + computed analysis figures.