Six Agibot humanoid robots underwent a public stress test involving 24 consecutive hours of real work across all positions at a normally operating convenience store in Xuhui District, Shanghai, assessed according to actual store standards. In the end, four robots successfully passed probation, while two were not converted to full-time status due to minor shortcomings in on-the-job initiative and position stability. The test directly confronted technical challenges in real convenience store scenarios, including fine manipulation of deformable objects, coupled long-horizon task planning, and obstacle avoidance in open dynamic environments. Leveraging an embodied Agent framework called Genie Master, the robots can autonomously complete error attribution, strategy adjustment, and experience accumulation, enabling them to iterate skills while on the job. This verified the feasibility of an embodied intelligence autonomous iteration path that does not rely on human fallback, and also reflected the industry characteristic that embodied intelligence deployment requires full-stack technical collaboration. The public test results visually demonstrated the actual gap between current humanoid robots and the requirements of real positions.