Prompting

Convert the Brief into a Testable Build Prompt

A nine-part prompt structure for clearer, safer AI-assisted builds.

Plan stage · Module 7 of 15

Outcome for this lesson

Turn the one-page brief into a compact specification with explicit behavior, evidence, and boundaries.

Bring forward: Bring the completed one-page brief from Module 6.

Produce: A reusable nine-part prompt plus acceptance-test list.

0 of 15 core modules complete

Progress is saved on this browser. Create a free membership to sync across devices.

Recommended prerequisite: complete Module 6, “Turn the Validated Idea into a One-Page Build Brief,” and bring its artifact. You may still use this page as a reference.

Plan stage

Build the next result

A prompt should implement the brief, not replace it. This module turns each approved decision into explicit instructions and acceptance tests that can be reused when the conversation changes.

Your result for this module

A reusable nine-part prompt plus acceptance-test list.

Continuous case study

Worked example: Prompt excerpt

“Build only add-and-display for the inquiry tracker. Use persistent labels and keyboard-operable controls. Reject invalid dates without clearing valid fields. Store no message content. Return changed files, test steps, known limits, and rollback instructions.”

Do the work

  1. Translate all nine sections

    Write goal, user, inputs, rules, outputs, interface, error handling, tests, and definition of done.

  2. Add boundaries and non-goals

    State prohibited data, excluded features, unsupported use, and decisions the model must not make.

  3. Provide representative examples

    Include ordinary, edge, invalid, accessibility, failure, and recovery cases.

  4. Request evidence, not confidence

    Ask for changed files, commands or checks run, results, assumptions, known limits, and rollback steps.

Copy-and-complete template

Create your module artifact

Goal:

User:

Inputs:

Rules:

Outputs:

Interface and accessibility:

Error and recovery behavior:

Tests and examples:

Definition of done and evidence:

Non-goals and prohibited data:

Keep sensitive information, passwords, API keys, payment data, and private customer details out of course notes and AI prompts.

Quality gate

Check the result before the quiz

  • The prompt traces back to the brief.
  • Error and recovery behavior are explicit.
  • Tests include more than one successful example.
  • The requested response contains verifiable evidence.

How this moves the project forward

Module 8 uses this package to build and preserve the first useful end-to-end slice.

Evidence checkpoint and lesson summary

Confirm the artifact, then answer the key questions

The questions cover a core idea, a realistic decision, and evidence from the lesson. Incorrect answers identify concepts to review; they are not a complete measure of your experience or ability.

Artifact checkpoint

This is an honest self-check. Do not enter private data here; keep the artifact in your own approved workspace.

1. What is the prompt blueprint supposed to accomplish?
2. What should the prompt say about invalid or unavailable data?
3. Which prompt package is ready to use?

Confirm the artifact and answer all three questions correctly to unlock the recommended next step.

About the author

Son Kim

Son Kim writes from first-hand experience as a blind AI user. He tests vibe-coding tools, models, plugins, connected hardware, and everyday AI workflows to document what improves accessibility, independence, productivity, earning opportunities, problem-solving, and quality of life—and what still requires human judgment.

Read the full author background

Review the editorial, correction, advertising, and affiliate standards