Everything an agent does starts with a prompt — the system prompt that defines its role, the instructions it's given, and the way tool results get fed back into it. Prompt engineering is the practical skill of shaping that text so the model does what you actually intend, reliably, instead of what it merely might.
This module starts with the fundamentals: what prompt engineering actually is and why small wording changes produce such different results. It then works through the core toolkit — zero-shot, few-shot, and chain-of-thought prompting — first as an overview of the main methods, then as a closer look at how and when each one actually helps. From there it turns to system prompts specifically, since for an agent the system prompt is doing most of the real work of defining its behavior. It closes with a full, hands-on course that puts all of it into practice against a real API.
Watch the lessons in order, then continue on to Model Context Protocol.
What prompt engineering actually is and why it matters, with real-world examples and a step-by-step look at how to craft a prompt that gets you the output you're after.
A tour of the main prompting techniques — zero-shot, few-shot, chain-of-thought, and more — and a practical sense of which one to reach for depending on the task in front of you.
A closer, example-driven look at the three techniques that do the most heavy lifting in practice: asking with no examples, asking with a few examples, and asking the model to reason step by step before answering.
Why the system prompt is the single most important lever for an AI agent specifically — how it shapes tool use, tone, and decision-making, and what separates a system prompt that actually works from one that just sounds good.
The full DeepLearning.AI & OpenAI course, taught by Isa Fulford and Andrew Ng: writing clear and specific instructions, giving the model time to think, and using prompting for summarizing, inferring, transforming, and expanding text — with hands-on API examples throughout.