This repository provides the complete reproduction package for a study demonstrating that observational systems carry structural skeletons — minimal, model-independent event subsets recoverable through an information-theoretic criterion — and that these skeletons are sufficient to recover the mathematical form of governing physical laws. CONTENTS - Full Python code (~2,300 lines) for six experiments - All datasets: MPC asteroid observations, NOAA buoy and climate indices, California Housing, synthetic oscillator data, exoplanet parameters - Detailed reproduction guides (experiment_1 through experiment_6) - Analysis outputs and results KEY FINDINGS REPRODUCIBLE WITH THIS PACKAGE - Recovery of Kepler's third law (p = 3.007 ± 0.008, R² = 0.99979) and the inverse-square law (ΔAIC = 95.7, 100% bootstrap) from raw telescope data, without any physics prior - Cross-scale Kepler verification across 2,887 bodies (0.004–1,031 AU) - Detection of hidden physical structure where classical statistics are blind (100% vs. 20%, d = 2.68) - Deterministic skeleton convergence (H = 0) across 8,000 bootstrap subsets - Cross-architecture transfer across 10 architectures from 4 model families TECHNICAL REQUIREMENTS Python 3.9+, NumPy, SciPy, scikit-learn. The entire computational pipeline completes in minutes on consumer hardware. The framework (ASE: Adaptive Skeleton by Elimination) embeds observational events in a five-dimensional geometric manifold and identifies structural invariants through an information-theoretic dual-threshold criterion requiring only input-output queries — no model-internal access needed.