Prompt engineering
The practice of crafting LLM input — instructions, examples, formatting, role assignments — to elicit reliable, structured, on-task outputs. Includes system prompts, few-shot examples, chain-of-thought scaffolding, and output schemas.
Background
Prompt engineering remains a real discipline despite increasing model capability. Effective prompts establish role (system prompt), constraints (length, format, tone), examples (few-shot), and explicit chain-of-thought scaffolding for complex reasoning. Production AI applications layer prompt engineering with structured output (JSON schema, tool use) and evaluations (golden test sets). The practice has shifted from "clever wording" to "specification-style writing" as models have improved at instruction following.