Prompt engineering
Writing and testing prompts to get better, more reliable answers from a model.
Prompt engineering covers practical techniques: giving clear instructions, supplying examples, asking for a fixed output format, breaking a task into steps, assigning a role, and telling the model what to do when unsure. It was especially important with earlier models, which were sensitive to phrasing. As models have improved at following plain requests, the emphasis has shifted to supplying the right context, sometimes called context engineering, and to testing prompts systematically against sets of examples.