Copied to clipboard!
Module 2 Foundations beginner 34 min

The anatomy of a real prompt

What you'll be able to do

  • Name the five parts of a prompt: role, task, constraints, examples, and output spec
  • Tell a prompt that "looks smart" from one that actually ships usable work
  • Rewrite a vague one-liner into a structured prompt and feel the quality jump

Rebuild a Vague Prompt Into a Real One

You type a lazy one-liner at the AI, get back something generic, and blame the model. The problem is almost never the model. A real prompt has five parts: role, task, constraints, examples, and an output spec. In this exercise you take a vague request you would actually type and rebuild it into a structured prompt, then watch the quality jump. You keep judgment over what "good" looks like; the AI just helps you see the parts.

Recommended tool

claude

Claude is strong at structured, multi-part work and at explaining its own reasoning, so it is a good place to learn the anatomy of a prompt and feel the difference a clear output spec makes.

Bring to the exercise

  • One vague one-liner you would honestly type on a busy day (for example, "write me a follow-up email")
  • A rough sense of who the output is for and what done looks like
  • Your AI Context Document, if you have one yet
Step 1

Start With the Lazy Version

Capture the prompt you would really type, unimproved, so you have an honest before-and-after. Do not fix it yet.

What to substitute before pasting

  • [PASTE A VAGUE ONE-LINER YOU WOULD ACTUALLY TYPE] A real lazy request, not a cleaned-up one. The messier it is, the more you will learn.
Here is a weak prompt I might lazily type on a busy day:

"[PASTE A VAGUE ONE-LINER YOU WOULD ACTUALLY TYPE]"

Do not answer it yet. First, tell me in plain language what is missing that would force you to guess, and where you would most likely guess wrong.
What good output looks like
  • Names the specific things it would have to guess (audience, length, format, facts).
  • Points to where a guess would most likely miss the mark.
  • Does not yet produce the deliverable.

Verify before using AI's output

Step 2

Label the Five Parts

Rebuild the prompt with all five parts named out loud, so you can see the anatomy instead of memorizing magic phrases.

Now rewrite my weak prompt with all five parts labeled:

ROLE: who you should be when you answer
TASK: the one job, stated plainly
CONSTRAINTS: the rules (use only facts I give you, length, format, tone, ask before assuming)
EXAMPLES: tell me exactly what I should paste so you can match my style
OUTPUT SPEC: the precise shape of a finished answer

After each part, give me one line on why it raises the odds of a usable result.
What good output looks like
  • All five parts are present and clearly labeled.
  • The role is specific leverage (a named point of view), not flattery like "act as an expert."
  • The output spec describes a concrete shape you could check against.

Verify before using AI's output

Step 3

Run Both and Compare

Feel the difference. Run the lazy version and the rebuilt version and judge which one actually ships.

Run two answers so I can compare:

A) Answer my original weak prompt exactly as written.
B) Answer the rebuilt five-part prompt.

Then tell me, in plain terms, what the rebuilt version got that the lazy one missed, and which single part of the five made the biggest difference.
What good output looks like
  • Version B is noticeably more usable and on-format than version A.
  • It names which part (often the output spec) moved quality the most.
  • The comparison is concrete, not "B is just better."

Verify before using AI's output

Step 4

Save It to Your Prompt Drawer

Turn the win into a reusable asset. A prompt that worked once is worth keeping with the changing parts marked.

Take the rebuilt prompt and turn it into a reusable template.

Mark every part that changes each time with a clear [BRACKET] placeholder. Keep the role, constraints, and output spec fixed. Then show me one filled-in example so I can see it run.

This goes into my personal prompt drawer, so make it clean enough to paste again next week without editing the structure.
What good output looks like
  • The reusable version uses bracketed placeholders only for the parts that truly change.
  • The fixed scaffolding (role, constraints, output spec) stays put.
  • A filled-in example proves it still works.

Verify before using AI's output