The brilliant new colleague

You met five prompting habits at the end of the first course. This course turns each one into a tool you can use on purpose, and you'll run every one live on Claude.

Start with the habit under all the others. Anthropic's advice is to treat Claude like a brilliant colleague on their first day. They're sharp and fast, but they know nothing about your project, your users or what “good” means to you. Anything you leave out, they have to guess.

Here's the test. Before you send a prompt, picture handing it to a smart person who has never seen your work. Would they need to ask you questions? If so, Claude will have to guess the answers, and guesses come out generic.

Same task, two prompts. The left one is “Write a description of our water bottle.” The right one adds who reads it, the facts, where it goes and how long it should be.

The left reply isn't wrong. It's just the average of every bottle blurb ever written, because that's all the prompt asked for. The right one could only be about this bottle.

Two more things about how current Claude models read you.

They take you at your word. Recent models follow instructions closely, and some very literally. Ask for a description and you get one description. If you also want a headline, three options or a bolder tone, say so. Claude won't assume you wanted extra.

Order matters, so number it. When the steps have to happen in a set order, or none can be skipped, write them as numbered steps instead of one long sentence.

Your turn. The draft is the vague prompt. Rewrite it so a new colleague could nail it first time: say who reads it, give real facts (make them up, it's practice), say where it will appear and set a length.

Run it a couple of times. The wording changes from run to run, but the facts and the shape should hold, because you pinned them down.