Abstract: With the extension of artificial intelligence technology from virtual interaction to entity interaction, the fusion of embodied intelligence and generative artificial intelligence has given birth to OpenClaw-like agents, which have full-link automation capabilities for environment perception, autonomous decision-making, and entity execution, and automated behavior has penetrated into all fields of social production and life. However, due to the lack of value alignment mechanism, agents also bring systemic risks. These risks run through the whole process of automatic behavior: legal deviation of value coding in goal setting link, improper compliance of goal disassembly; legal misjudgment of scene intention in intention understanding link, legal misinterpretation of human instructions; behavior anomie of physical execution in behavior execution link, boundary encroachment of authority exercise; adaptation defect of dynamic scene in value adaptation link, coordination disorder of multiple values. Therefore, the legal regulation of agent automation behavior should take value alignment as the main line, and construct a full-chain regulation system based on its specific and generative characteristics, which includes pre-regulation, in-process control, and post-accountability. Specifically, compliance guidance shall be given to goal setting, precise adaptation of intention understanding, boundary constraint on behavior execution, dynamic adjustment of value adaptation, value orientation, and mandatory requirements of legal norms shall be embedded in each link of intelligent automation behavior, so as to provide legal guarantees for compliance application of intelligent bodies and high-quality development of intelligent economies.