Holders can write prompts that reliably produce usable results, applying the GCSE framework along with chain-of-thought, few-shot prompting and grounding. They can diagnose why a prompt failed and fix it, and evaluate an AI output before acting on it.