Definition
Prompt engineering covers clear task instructions, the role the model should take, worked examples (few-shot prompting), the format of the output, and rules for edge cases. Small changes in wording can change results, so prompts are tested and iterated like code.
As systems grew into multi-step agents, the work broadened into context engineering: managing everything the model sees, not only the instruction. Prompts remain the cheapest lever to pull before retrieval, fine-tuning or a larger model.