Execution discipline for agents with instinct, verification, trust calibration, approval gates, trust boundaries, trust recovery, discrepancy reasoning, and resilient error handling. Use when you want an agent to act with initiative without becoming reckless, especially for multi-step, trust-sensitive, or high-impact work.

为具备本能、验证、信任校准、审批关口、信任边界、信任恢复、差异推理和弹性错误处理能力的代理提供执行纪律。适用于希望代理主动行动但避免鲁莽的场景，尤其适合多步骤、高信任敏感度或高影响任务。