Section 2: Prompting & Iteration

If Section 1 was about understanding what Claude is, Section 2 is about learning how to talk to it well. The gap between a mediocre Claude response and a genuinely useful one is almost always in the prompt, not the model. This section teaches you to close that gap deliberately and consistently.

Why this section matters

Most people write prompts the way they write a text message: quickly, loosely, with the assumption that the other party will fill in the gaps. That works with people who share your context and can ask follow-up questions. With Claude, vague inputs produce generic outputs, and generic outputs require more back-and-forth to fix than a well-structured prompt would have cost upfront.

Good prompting isn’t a technical skill. It’s a communication skill. The underlying principles are the same ones that make any written brief effective: be clear about who you are, what you need, what context is relevant, and what good looks like.

What you’ll learn

This section covers five topics:

Role, task, context, and output format. The four-part structure that turns a vague request into a clear brief. Role tells Claude who it’s working for and what perspective to take. Task specifies what you want done. Context provides the relevant background. Output format tells Claude what the finished product should look like. Mastering this structure improves every prompt you write.

Adding examples and constraints. The fastest way to align Claude’s output with your expectations is to show it what you want. Examples anchor tone, structure, and quality in a way that descriptions alone rarely achieve. Constraints (what to avoid, what to keep under a certain length, what format not to use) are equally important: they prevent Claude from doing plausible things you don’t want.

Asking for tables, outlines, and checklists. Structured output requests are underused. When you need information organized rather than narrated, asking for a specific format produces a much more usable result.

Using follow-up questions. A single prompt rarely produces a finished result. Knowing how to refine, redirect, and build on Claude’s initial response through follow-up is how you move from first draft to something genuinely useful.

Critique and revision loops. Claude can critique its own output and improve it on request. Building a deliberate revision loop into your workflow produces better results than accepting the first response or rewriting manually from scratch.

The mini project

You’ll take a weak, poorly-specified prompt (one you’ve actually used or one provided as an exercise) and improve it through each element: role, context, examples, output format, and a revision pass. The before-and-after comparison is the deliverable, and it’s a practical demonstration of exactly how much prompt quality affects output quality.

How to approach this section

Each topic builds on the previous one. Start with the four-part structure, then layer in examples and constraints, then practice requesting specific formats, then work on the follow-up and revision skills. By the end, prompting well should feel less like trial-and-error and more like a learnable, repeatable process.

The payoff compounds. Every Claude interaction in the rest of this course benefits from stronger prompting habits.