Back to course overview
Module 2Prompting basics 18 min

Lab: Prompt practice

Three graded drills — brief, imitate, iterate — on real tasks from your own week, ending with your personal prompt template.

Three drills, each on a real task from your own work. By the end you'll have a personal prompt template you'll use daily — and reuse in the capstone.

Drill 1 — The five-part brief

  1. 1Pick a document you actually owe someone this week (email, summary, plan, announcement).
  2. 2Write a prompt with all five parts: role, task, facts, constraints, format. Label them while practicing.
  3. 3Run it. Grade the output A–F before editing anything.
  4. 4If below A: identify which of the five parts was weakest, strengthen only that, re-run. Repeat once more.

Drill 2 — Style transfer

  1. 1Find two examples of a format your team produces often (status updates, tickets, customer replies).
  2. 2Paste both and ask for a new one 'in exactly this style' from fresh facts.
  3. 3Show the result to a colleague without saying which was AI-written. Ask them to spot it. Debrief what gave it away — that gap is your style guide. Working solo? Paste both versions into a fresh chat and ask the model to pick the AI-written one, then compare with your own read.

Drill 3 — The editing conversation

  1. 1Take Drill 1's output. Run the self-critique prompt from the last lesson.
  2. 2Give two editor-style directions ('move the ask up', 'cut the throat-clearing').
  3. 3Ask for three tonal variants of the final version and pick one.
  4. 4Total time check: the whole loop should be under ten minutes. If it isn't, your first brief was underspecified.

Deliverable: your template

my-prompt-template.txttext
You are [ROLE relevant to the task].
Task: [ONE VERB + object].

Facts you need:
- [fact]
- [fact]
- [paste source material below the line]

Constraints: [length] · [tone] · [audience] · avoid [X].
Format: [exact shape of the output].

Before answering: list any information you're missing that would change the answer.
That last line is doing real work

"List any information you're missing" flips the failure mode from Module 1 — instead of silently guessing, the model asks. Keep that line in every serious prompt you write.

Problem set 2

Each of these five prompts is underspecified — a different one of the five parts is the main gap in each. For each: diagnose which part is missing, rewrite the prompt to fix it, then run both versions and check whether the output changed the way you predicted.

  1. "Write an email to the team about the deadline."
  2. "Here's our pricing page copy: [paste]. Thoughts?"
  3. "You are a marketing expert. Write a product description for our new product."
  4. "Summarize last quarter's customer feedback: [paste]. Include everything important."
  5. "Draft the reorg announcement. Facts: support and success teams merge under Dana on May 1; no role changes. Keep it professional."
PreviousIterating on outputNext Productivity use cases