Prompt Engineering

Prompt engineering is about guiding LLM behavior without changing model weights: role setup ("You are a legal advisor"), few-shot examples, and strict output requirements (for example, JSON schemas). Strong prompts reduce hallucination.

Core related terms: prompt, context window, and hallucination.

---

Key characteristics

• Guides model behavior through instruction design, structure, examples, and constraints.
• Improves precision in many everyday cases without requiring model fine-tuning.
• Quality depends on clear task definition, output format constraints, context quality, and boundaries.