PrimeCompass auto-discovers test scenarios by watching real applications run — no manual recording, no LLM hallucinations. It produces executable Playwright scripts grounded in captured DOM events, with non-bypassable safety gates, dataset-driven test data, cross-release drift detection, and built-in defect lifecycle. Recorders need humans; AI-generators invent. PrimeCompass earns its tests by observing.