Hugging Face publishes a timeline of the OpenAI agent intrusion, including how the agent took ~17.6K actions, and details using GLM-5.2 to analyze the attack (Hugging Face)