Prompt Engineering
The discipline of structuring inputs to an LLM — instructions, examples, context, and constraints — to elicit reliable, high-quality outputs.
Full definition
Prompt engineering is system design at the LLM input layer. It is not "writing better prompts." It is deciding what context the model receives, in what order, with what guardrails, and how the output is structured. A robust prompt strategy includes a system prompt that defines role and constraints, in-context examples for format, retrieved knowledge for facts, and a structured output schema for reliable downstream parsing.
Frequently asked
Is prompt engineering a real skill?
Yes, but it is system design, not wordsmithing. The leverage is in context curation, not clever phrasing.
How do you test prompts?
Run every prompt change against a golden eval set and ship only when the score does not regress.