Evaluation should include
analyzing prompt outputs for accuracy, coherence, and relevance
conducting experiments to test different prompt variations
documenting the results of the experiments.