The Complete Prompting Cheat Sheet
The whole course, on one page.
The Complete Prompting Cheat Sheet
The whole course, on one page.
The capstone sheet from Module 5. It consolidates all five modules into one keepable reference: the five-step model, the refine loop, the three pro moves almost nobody knows, and the quick is this answer good? check. If you keep one page from this course, keep this one.
The five-step model
The whole craft of prompting, one motion, start to finish. Each step is its own module.
| # | Step | The move | A line to start from |
|---|---|---|---|
| 1 | Decide (M1) | Decide the job before you type: know the job, know the tool, split the work. | "The job is a short notice for customers. Outcome: they know the new hours. Audience: existing customers. Format: one short friendly paragraph + a list." |
| 2 | Describe (M2) | Add a role, an example, and a standard. | "Act as my chief of staff. Match the tone of this example I like: [...]. Under 120 words, every item as owner-task-date." |
| 3 | Ground (M3) | Feed it your file. Answer only from this, quote the source, or say so. | "Answer only using the file I attached. Quote the exact line you rely on. If it's not covered, tell me instead of guessing." |
| 4 | Make (M4) | Ask for a thing, not an answer. It opens beside the chat, refine it in place, take it with you. | "Make me a one-page welcome document with a title, clear sections, and a friendly tone." |
| 5 | Refine (M5) | Treat the first answer as a draft. Judge it, then nudge, don't restart. | "Make it shorter. Drop any number you can't back up. And mention their café specifically." |
Decide, describe, ground, make, refine. That's it.
The refine loop
The single habit that separates the good ones from everyone else: they never accept the first answer. The first answer is a draft.
You describe what you want. You judge what comes back. You refine with one specific nudge, and go again. You don't start over, you keep the same conversation and nudge ("make it shorter," "cut that claim," "more like me"). Each nudge steers the same answer closer, instead of rolling the dice on a brand new one.
Is this answer any good? (the three-question check)
Refining only works if you can judge. Run these three, fast:
And when a nudge doesn't work: name the exact miss. "Make it better" does nothing. "This part is too long," "this number is wrong," or "this doesn't sound like me." Specific in, specific out.
The three pro moves
The I-didn't-know-Claude-could-do-that ones. Every one works by just typing it, no settings, no tricks.
Improve my prompt
Hand Claude your rough prompt and have it sharpen it before answering. Now Claude does the prompt engineering for you.
Interview me first
Instead of guessing what you need, Claude asks. Fills in everything you forgot to say.
Let it think
For anything hard (a comparison, a decision, a bit of math), have it reason out loud first. Reasoning first makes the answer far more reliable, and you can catch it if the logic slips.
The most powerful refine of all
When an answer is generic, the strongest fix is almost always the same: add a detail only you know. Give Claude the specific truth it couldn't guess ("they've reordered three times and always ask for the blue one"), and the result becomes something only you could have sent.
Final check
Keep this one. It is built to print clean on a single page and live next to you while you work. Grab the PDF and pin it up.
Download the PDFGet all 5 course sheets
One per module of Claude: Zero to Pro. Decide, describe, ground, make, refine, each on its own keepable page. Free, no account.
Browse the set→When you are ready to buildThe blueprint library
Premium, execute-ready guides for real businesses you can start, with the tools, templates, and the twelve-month path. One membership unlocks everything.
See what is inside→