Prompt engineering courses charge $2,000 to teach what this article does for free! Most people using Claude are still playing the prompt lottery instead. Type something in. Hope something useful comes out. Add more words when it is not right. Start over with slightly different phrasing. Repeat until something usable appears or you give up. There is no secret. There is a foundation. → identity: tell Claude who it is for your specific work — role assignment is the single highest-leverage change → context: give Claude the minimum required to produce excellent work, not just to understand the task → standards: define what good looks like and what Claude should never do → write all three once — every session loads from them automatically Once you understand the system, the outcomes stop feeling random. You know exactly why an output was good and exactly how to fix one that was not. Follow @neil_xbt for more Claude foundations content that shows you what consistently excellent results actually require.