Most prompts improve through iteration: run the prompt on real inputs, look at where the output goes wrong, change one thing and run it again. Doing that deliberately, with a fixed set of test inputs and a record of each version, turns guesswork into measurable progress.
Automatic prompt optimization takes the same loop further and lets a model propose and score new prompt versions, as in APE and ProTeGi or with frameworks such as DSPy. These methods need a good evaluation set; without one they optimize toward the wrong thing. The guides here cover refining a prompt by hand in a few rounds, diagnosing bad outputs to find what to fix, and automatic optimization methods with their costs and limits.
The library lists all guides with filters for topic, level, category and reading time.