Master the Art of Prompt Engineering

What is GPT?

GPT stands for Generative Pretrained Transformer—a type of AI model trained on large amounts of text to generate helpful responses, write content, answer questions, and more.

Anatomy of a Good Prompt

A strong prompt is usually built from a few clear parts:

Introduction

Sets the context and intent (what you’re trying to do).

Primary Content

The core instruction or question.

Simple prompt: one clear instruction

Complex prompt: multiple constraints, steps, or requirements

Examples (Shots)

Examples guide the model toward the style and structure you want.

Zero-shot: no examples

One-shot: one example response

Two-shot (or few-shot): two or more examples

Cue (Helpful Context)

Extra information that improves accuracy and relevance—especially for recommendations.

Example: location, budget, dietary restrictions, preferences, etc. Cues can be:

Zero cue: no extra context

One cue: one key piece of context

Multiple cues: several details for higher-quality answers

Support Content

Any additional details like constraints, formatting requirements, tone, audience, length, and “do/don’t” rules.

Types of Prompts

Different prompt styles work best for different tasks:

Chained prompts: multiple steps where each response feeds into the next Example: brainstorm → outline → write → edit

Conditional prompts: include a condition that changes the answer Example: “If it’s raining outside, what should someone wear?”

Open-ended prompts: allow free-form responses without strict constraints

Structured prompts: demand a specific format (bullets, tables, JSON, sections, etc.)

Useful Action Words (Prompt Verbs)

Clear verbs make instructions easier to follow and outputs more consistent:

Analyze, Define, Outline, Suggest, Arrange, Create, Explain, Rephrase, Clarify, Differentiate, List, Rewrite, Combine, Discuss, Recommend, Summarize

Key takeaway

Prompt engineering is mostly about being clear (what you want), specific (constraints and format), and guided (examples and cues). The more you shape the input, the more predictable—and useful—the output becomes.