Don't write complex loop prompts manually every time.

You just need to clearly state the goal, background, constraints, and success criteria, and let the Agent generate a set of executable task plans on its own.
In the past, humans kept writing prompts to command AI.
Now, you first let AI help you break down tasks, set boundaries, and establish checkpoints, then let it execute according to the plan.
Simply put, what really matters is not the prompt