SKILL-ieee-rl-dl-ml-reproducible-training

Use when RL/DRL/DL/ML paper reproduction or energy-system control work needs unattended long-goal execution, long training, full-resume checkpoints, mandatory TensorBoard long-run visualization when applicable, traceable train/test evaluation, same-tier raw environment reward delta comparisons when applicable, online-checkable CSV/JSONL/stdout records, IEEE TSG/TPS figures, Chinese-first colleague-readable reports, post-training PPT/report packs, figure/table quality audit, SHA256 manifests, artifact indexes, risk analysis, or cleanup/exclusion of failed or half-finished artifacts.