Lesson 3 of 11 in the free starter module · Progress is saved in this browser

All 11 lessons
  1. 1 What an AI chatbot actually is
  2. 2 Give it enough context to help you
  3. 3 Prompting Lab 1 — The Prompt Card
  4. 4 Prompting Lab 2 — Before and after
  5. 5 Improve the answer instead of starting over
  6. 6 Prompting Lab 3 — Fix-this-prompt drills
  7. 7 Prompting Lab 4 — Your personal prompt template
  8. 8 Check before you trust
  9. 9 Keep private things private
  10. 10 Run the full loop on a real task
  11. 11 Prompt Practice Lab and starter mastery check

Free starter · Prompting Lab · Lesson 3

Prompting Lab 1 — The Prompt Card

Time
about 25–30 minutes
Prompting skill
P1 Clear ask, P2 Context, P3 Reader and tone, P4 Shape and limits; light previews of P8 Role and P9 Show an example; first look at P6 and P7
You'll need
Your free AI chatbot, a new chat, and paper or a notes app.

Listen Talk to your coach


1. Objective

By the end of this lesson, you will be able to:

  • Name the parts of a strong prompt on the Prompt Card. (Remember)
  • Label the parts in a prompt someone else wrote, and spot the missing ones. (Analyze)
  • Write a prompt for an everyday job that scores at least 1 on rubric rows 1–4. (Apply)

2. Hook

Listen: welcome 0:12 · AI voice: Bella
Transcript

Welcome to the Prompting Lab. Over the next few lessons, you'll take prompts apart, put them back together, fix broken ones, and leave with a template that's yours. Today we start with the parts.

A recipe card has the same lines every time: ingredients, amounts, steps, oven temperature, time. You don't fill in every line for a salad. But because you know the lines, you notice right away when one that matters is missing.

A prompt works the same way. This lesson gives you the card.

3. Warm-up (from lessons 1 and 2)

  1. (Lesson 1) What are the two halves of a clear ask?
  2. (Lesson 2) Name the four questions to glance over before you send a prompt.
  3. (Lesson 1) Why can a smooth, confident reply still be wrong?
Answers
  1. What you want done, and what you want back (the task and the shape of the result).
  2. Who is this for? Why am I asking? What tone? What limits?
  3. The model predicts likely next words. It's very good at fluent writing, and fluency isn't the same as truth.

4. The idea

Listen: the key idea 0:44 · AI voice: Bella
Transcript

Meet the Prompt Card. It's a checklist, not a form. Row one is the task: what you want done, and what you want back. Row two is context: the situation and the facts it must use. Row three is the reader and the tone. Row four is shape and limits: format, length, what to include, and what to avoid. Two optional extras can help: a role, which changes tone and focus but not truth, and a short example of what good looks like. Then two safety rows. Row five, safe to send: nothing private goes in. Row six, a check built in: when facts matter, ask it to say if it's not sure. Always give the task, and always keep it safe to send. Add the rest when it helps.

The Prompt Card

The Prompt Card is the checklist you'll use for the rest of this course. Its rows match the Prompt Check, the 0–2 self-grading rubric you met in lessons 1 and 2.

Row Part The question it answers Example words
1 Task What do I want done, and what do I want back? "Write…", "Explain…", "Make a list of…"
2 Context What's the situation? What facts must it use? "My neighbor is 80 and has never made a video call…"
3 Reader and tone Who is it for, and how should it sound? "for a beginner," "warm, not jokey"
4 Shape and limits What format and length? What to include or avoid? "a numbered list of 7 steps," "no jargon"
extra Role (optional, preview) Who should the chatbot act as? "Act as a patient tech helper."
extra Example (optional, preview) What does good look like? "Write steps like this: '3. Tap the green button.'"
5 Safe to send Is everything here fine for a stranger at that company to read? placeholders like [ACCOUNT NUMBER]
6 Check built in If it matters, did I ask it to flag uncertainty or name what to check? "If you're not sure, say so."

On the back of the card, a memory line:

Say what you want, about what, for whom, in what shape — and, when it helps, as whom and like what. Then ask: is it safe to send, and how will I check?

Listen: the memory line 0:18 · AI voice: Bella
Transcript

Here's the memory line one more time, slowly: say what you want, about what, for whom, in what shape — and, when it helps, as whom and like what. Then ask: is it safe to send, and how will I check? Try saying it back before you read on.

Google's own prompting guide for Gemini describes four similar areas — persona, task, context, and format — and notes you don't need all four every time; OpenAI and Anthropic give similar advice about being clear, giving context, and showing examples (see Sources). The names on the card are ours, chosen to be easy to remember.

New words, in plain language

  • The task is the job itself. It usually starts with a verb. Google's guide calls the verb the most important part of a prompt.
  • A role is a part you ask the chatbot to play. It shapes tone and focus. It does not give the chatbot real expertise or credentials, and it doesn't make facts more reliable. You'll practice roles properly in Module 4.
  • An example is a short sample of the kind of thing you want, so the chatbot can match the pattern. (Technical guides call this "few-shot" prompting.) Module 4 goes deeper.
  • Shape and limits (you met these as constraints in lesson 2) are the format, length, and rules: what to include, what to avoid.

Rows 5 and 6: a first look

You'll go deeper on these in lessons 8 and 9. For now, two simple habits:

  • Row 5, Safe to send: leave out passwords, ID numbers, account numbers, health records, and other people's private details. Use a placeholder like [NAME] instead.
  • Row 6, Check built in: when a fact matters, add a line like "If you're not sure about something, say so."

You don't need every row every time

  • Always: row 1 (task) and row 5 (safe to send).
  • Almost always: rows 2 and 4.
  • When it helps: row 3, role, example, and row 6.

More words aren't automatically better. A prompt stuffed with unnecessary instructions is harder for the model to follow and harder for you to fix. The aim is the right information.

5. Worked example

Here's a prompt that uses every part of the card. Read it, then read the notes.

Act as a patient tech helper.                                   ← ROLE (extra)
Write step-by-step instructions for making a video call         ← ROW 1: TASK
on a smartphone.
They're for my neighbor, who is 80, has an Android phone,       ← ROW 2: CONTEXT
and has never made a video call. She wants to call her          
grandson, who's already in her contacts.
Write for a complete beginner, in a calm, reassuring tone.      ← ROW 3: READER AND TONE
Use a numbered list of no more than 7 short steps, and add      ← ROW 4: SHAPE AND LIMITS
one encouraging sentence at the end. No technical terms.
Write steps like this: "3. Tap the green phone button."         ← EXAMPLE (extra)
If a step might look different on her phone, say so.            ← ROW 6: CHECK BUILT IN

Why each part is there:

Part What it prevents
Role Steps that are correct but brisk
Task A history of video calling instead of instructions
Context Guessing iPhone instead of Android, or assuming she knows the basics
Reader and tone Jargon, or a rushed tone
Shape and limits A wall of text she can't follow
Example Long, multi-action steps
Check built in Steps that assume every phone's screen looks the same

Row 5, Safe to send: notice what's not there — her name, phone number, address, or health details. The chatbot doesn't need them. Score: 2.

Listen: example conversation: Rosa asks about the Prompt Card 0:16 · AI voices: George, Jessica
Transcript

Rosa (a learner): Wait, do I really have to write all that every time?

Guide: Not at all. The card is a checklist, not a form. For a quick question, the task is enough. For something that matters, run down the card and add what's missing. You'll get faster with practice. That's what the labs are for.

6. Guided practice

Part A — Label the parts (paper, no chatbot)

Read this prompt and write which rows of the card you see, and which are missing.

I'm hosting a book club for six friends next Thursday. We're reading a mystery novel. Give me 8 discussion questions as a numbered list. Keep them open-ended and spoiler-free for the first half of the book.
Answer
  • Row 1, Task: "Give me 8 discussion questions."
  • Row 2, Context: book club, six friends, next Thursday, a mystery novel.
  • Row 4, Shape and limits: numbered list of 8; open-ended; spoiler-free for the first half.
  • Row 5, Safe to send: yes — nothing private. Score 2.
  • Missing: row 3 (reader and tone) is implied by "friends," which is fine here. No role or example, and neither is needed. Adding the book's title would be useful extra context.

Part B — Same job, two prompts

Start a new chat and send this.

Prompt 1 (copy and paste):

Plan a birthday party.

What to look for: probably a general plan, or questions back to you. Either is fine. The chatbot is doing its best with almost nothing.

Now open another new chat (so the first reply doesn't influence it) and send this.

Prompt 2 (copy and paste):

Act as a practical party planner.
Plan a 2-hour birthday party at home for my son's 7th birthday this Saturday afternoon.
Context: 10 kids, a small backyard, a moderate budget, and two adult helpers. He loves dinosaurs.
Format: a simple timeline in 15–30 minute blocks, then a short shopping list grouped by store section.
Limits: no activities that need special equipment, and no foods with nuts (one child has a nut allergy).

What to look for: a reply that's usually more specific to the age, theme, time, and allergy. Exact details will vary by tool and by try.

A note on safety: the allergy is a safety fact. The plan is a draft. For a real party, you'd still read food labels yourself and check with the child's parent. A chatbot can't confirm what's in a specific product.

Part C — Ask the chatbot to label your parts

In the same chat as Prompt 2, send this.

Prompt 3 (copy and paste):

Look at the prompt I sent you just before this one. Label which of these parts it included: task, context, reader and tone, shape and limits, role, example. Then name one part that's missing and suggest one sentence I could add for it. Keep your answer under 120 words.

What to look for: your parts labeled, probably with "example" or "reader and tone" named as missing. The suggestion is just a suggestion. You decide whether it helps.

7. Independent practice

Spot the weak prompt

You need help writing a short speech for a friend's retirement lunch. Which prompt is weakest?

  • A. Write a 2-minute retirement speech for my coworker of 15 years, Jo. Warm and a little funny, for an audience of about 30 coworkers. Mention her famous lemon bars and that she trained most of us.
  • B. Write a retirement speech.
  • C. Write a short, warm retirement speech for a coworker, about 2 minutes long.
Answer

B is weakest: row 1 is there (a task), but rows 2, 3, and 4 are all 0. C adds tone and length (rows 3 and 4 score 1–2) but no context. A scores well on rows 1–4. Using a first name in a speech draft is usually fine; if you'd rather not, use [NAME].

Fix-it drills

Drill 1 (with a hint).

Explain how to start composting.

Hint: Where do you live (apartment or house with a yard)? What shape should the answer take?

Drill 2 (no hint).

Give me ideas for dinner.
One strong version of each (not the only one)

Drill 1:

Explain how to start composting in a small apartment with no yard, for a complete beginner. Give me 5 short steps and a list of what can and can't go in. Plain language.

Drill 2:

Give me 3 weeknight dinner ideas that take 30 minutes or less, using chicken and rice I already have. One line each. Nothing spicy.

8. Rubric self-check

Score Prompt 2 from Part B, then your two drill fixes, with rows 1–6 (0 = missing, 1 = there but vague, 2 = clear and specific; "n/a" if a row doesn't apply).

Row Question Prompt 2 Drill 1 Drill 2
1 Task Does it say what I want done and back?
2 Context Did I give the situation and facts?
3 Reader and tone Who's it for, how should it sound?
4 Shape and limits Format, length, include/avoid?
5 Safe to send Only what's needed?
6 Check built in If it matters, did I ask it to flag uncertainty?

Reading your score: row 5 must be 2. After that, look at your lowest row, not the total. That's your next fix.

9. Retrieval check

  1. Name rows 1–4 of the Prompt Card.
  2. Which two parts are optional extras, previewed here and taught fully in Module 4?
  3. True or false: telling a chatbot "Act as a doctor" makes its medical facts reliable.
  4. You ask for "some ideas for dinner" and get 20 elaborate recipes. Which rows would most likely fix that?
  5. (From lesson 2) What does "context" mean?
Answers
  1. Task; context; reader and tone; shape and limits.
  2. Role and example.
  3. False. A role shapes tone and focus, not truth. Check anything that matters.
  4. Row 4, shape and limits ("3 ideas, one line each"), and row 2, context ("weeknight, 30 minutes, I have chicken and rice").
  5. The background the chatbot needs: the situation and the facts it must use, so it doesn't have to guess.

10. Recap

Listen: recap 0:28 · AI voice: Bella
Transcript

To recap the Prompt Card: rows one to four are task, context, reader and tone, and shape and limits. Role and example are optional extras. Rows five and six keep you safe: safe to send, and a check built in. Always give the task, and always keep it safe to send. A role changes tone and focus, not truth. This week, before you send an important prompt, run down the card once.

  • The Prompt Card: rows 1–4 (task, context, reader and tone, shape and limits), two optional extras (role, example), and two safety rows (5 safe to send, 6 check built in).
  • Memory line: say what you want, about what, for whom, in what shape — and, when it helps, as whom and like what. Then: safe to send? How will I check?
  • Always give the task, and always keep it safe to send. Add the rest when it helps.
  • A role changes tone and focus, not truth.

You can now: build a prompt on purpose, part by part, and spot what's missing. Use it this week: before you send an important prompt, run down the card once.

Next: lesson 4, Prompting Lab 2. You'll study before-and-after prompts, then write your own "after" versions.

11. Talk it through with your coach

  • Coach my prompt: read your Part B or drill prompt aloud. Your coach names the card rows it hears and asks one question about the weakest row. It gives a hint, not a rewrite.
  • Text-only alternative: sections 6–8 above.

Talk to your coach

River is an AI voice coach, not a person. River knows this lesson and can quiz you, listen to a prompt you're working on, and give hints. River won't do the exercise for you. You can talk out loud or type.

Sources

  • Google, "Prompting guide 101" for Gemini for Google Workspace (four main areas: persona, task, context, format; "You don't need to use all four in every prompt"; the verb is the most important part of the task), checked 22 November 2026: workspace.google.com/resources/ai/writing-effective-prompts; October 2024 PDF edition: services.google.com
  • OpenAI Help Center, "Prompt engineering best practices for ChatGPT" (be clear and specific, give context, refine iteratively, describe tone), checked 22 November 2026: help.openai.com
  • Anthropic, "Prompting best practices" (be clear and direct, add context, use examples, role prompting), checked 22 November 2026: platform.claude.com
  • Google, "Prompt design strategies" (clear instructions, examples, constraints, response format), checked 22 November 2026: ai.google.dev