OpenClaw: The High-Risk Security Trap of AI Assistants | KnowAI Space