Prompt engineering systematically designs and refines prompts to guide generative model outputs. An LLM cannot read your mind, so the prompt must communicate the intended result clearly. section 3.0 section 3.1
Treat it as both science and art: use theories and proven frameworks, then experiment, test, and adjust because natural language allows multiple effective approaches. section 3.0
A practical prompt-design process:
- Define the desired task and provide enough relevant context for a useful response. section 3.1
- Remove irrelevant detail as well as vagueness; extremely short prompts invite tangents, while overloaded prompts can obscure the command. section 3.1
- Account for the finite context window, since excessively long prompts can degrade response quality. section 3.1
- Experiment with prompt structures and templates, then evaluate outputs rather than assuming the first response is exceptional. section 3.0 section 3.1
A standard prompt template
A thoughtful-daughter scenario demonstrates how a structured prompt can turn a vague request into a more directed interaction. Role, task, features, and examples provide useful organizing elements. section 3.1
Figure 3.2 The standard prompt template in action—for a thoughtful daughter
Use the visual template example to see how a prompt moves beyond a generic request toward a specified task. section 3.1
