AI Prompt Engineering: write prompts that get sharp answers

AI Prompt Engineering is an interactive prompt engineering course that takes you from vague requests to prompts that get sharp, reliable answers, then on to briefing AI agents. Its 26 short sections cover tokens, the anatomy of a prompt, being clear and specific, few-shot examples, chain-of-thought reasoning, system prompts, output formats, follow-ups, agents and tools, prompt injection and multi-agent workflows.

It opens with a prompt builder: switch the role, context, examples and format on and off and watch the reply go from vague to sharp. That is the whole idea of the course in one demo.

What the course covers

Writing better prompts

Every prompt needs a task, and four optional parts make it sharper: a role, context, examples and the format you want. The course's rule of thumb is to write as if briefing a smart new colleague who knows nothing about your situation: who the result is for, what it is for and what good looks like. Examples (few-shot prompting) show the pattern you want, and asking the model to work through a problem step by step (chain of thought) helps on multi-step tasks.

Prompts for apps and agents

A system prompt sets standing instructions, a job, rules and tone, for a whole conversation, which is how apps give their AI a personality and limits. Agents go further: they take actions with tools such as web search or a code editor, choosing each tool from its name and description. So an agent prompt is a brief, with the goal, context, limits and how to tell the job is done.

Role:     You are a careful science teacher.
Context:  My class is 14 and has just met photosynthesis.
Task:     Write 5 quiz questions on what plants need to make glucose.
Format:   A numbered list, each with the answer in brackets.

Staying safe

Agents read web pages, emails and files, and some text may hide instructions meant to hijack them. This is prompt injection, and the course covers how to brief agents so they treat what they read as data, not orders.

Try a question from the course

Models read text as tokens, chunks of a few characters. For English, a common rule of thumb is that one token is about four characters, or three quarters of a word. So a 600-word email is about 600 ÷ 0.75 = 800 tokens. Our guide what is a token in AI explains why, with real splits and prices.

Now estimate a token count or a cost, then press Mark to check your answer in the course.

Who this course is for

• Anyone who uses ChatGPT, Claude, Gemini or Copilot and wants better answers first time.
• Developers and product builders writing system prompts and tool descriptions for AI features.
• Teams starting with AI agents, who need to brief them well and keep them safe.

What you need to know first

Nothing beyond having used an AI assistant. The token section uses simple division to estimate counts and costs. If you want to know what happens inside the model, the AI Models course explains neural networks and transformers.

How long it takes

There are 26 sections and most take a few minutes. Learners without a plan can open two sections of a course a day, in order, so the whole course takes 13 days. A plan removes the daily limit and lets you go straight to any topic, such as agents.

Read next

AI · How AI worksHow to write better AI prompts: 7 techniques that workHow to write better AI prompts: give context and a clear goal, show examples, ask for a format, let the model reason, and refine with follow-ups.