SCN Temporal-Entrainment Benchmark v1.0

A frozen, versioned, seed-reproducible phase-tracking benchmark for temporal / clock-inference systems: infer the environmental phase phi*(t) of a synthetic 24.0 h zeitgeber from its observable channels alone, across a 24.2 h-intrinsic / 24.0 h-zeitgeber detuning and a mid-run 6 h schedule shift. Ships v1.0/ (schedule.csv input, labels.json answer key, MANIFEST.json with SHA-256 pins), a standard-library-only scorer (score.py) whose metric definitions are the spec, and a reference solver using the accompanying SCN-LLM reference implementation. Metrics: raw phase error (primary, target-facing; includes the physical entrainment angle), entrainment angle, lock residual (locking tightness -- a separate metric, not a substitute for the phase-error target), and re-entrainment time. Synthetic data only; it measures phase-tracking, not believability, engagement, trust, or retrieval quality. Accompanies the SCN-LLM paper. The reference implementation is archived separately (DOI 10.5281/zenodo.21286077).