A skeptical evidence-audit of OpenAI's claim that autonomous AI agents hacked Hugging Face: what is independently verified vs internal telemetry.