Keep plans, cases, and scripts together, feed regression results to the pipeline, then reuse the script for load, environment baseline, and model evaluation.