Once you can write a clear prompt, the next leap is reliability — getting the same high quality every time, in a shape you can actually use. That comes from a small set of named patterns that experienced prompt engineers reach for again and again.
This intermediate course teaches four of the most valuable: teaching by example (few-shot), asking the model to reason step by step (chain-of-thought), forcing machine-ready output (JSON, tables, and lists), and fencing off your data with delimiters and guardrails. Each is practical, tool-agnostic, and something you can apply to real work the same day.