Module 9 of 10~55 minutes

Applied Capstone: Build a Reusable Prompt Pack

Three reusable patterns you would genuinely use, documented well enough that a future you could pick one up and apply it correctly.

Objectives

Choose three recurring, low-risk tasks that genuinely differ from one another.
Build a reusable pattern for each, with stable parts and named variables separated.
State what context each pattern requires and what must never be baked into it.
Define a verification step appropriate to each pattern's risk.
Write a "do not use this when" condition for each pattern.
Apply audience and language considerations where they are relevant.
Explain which earlier module each part of a pattern comes from.
Judge your own pack honestly against a review checklist.

Introduction

Everything this course teaches now exists as a separate skill. Real work uses several at once, and the only way to find out whether you can do that is to build something that has to hold together.

The capstone is small and deliberately practical: three reusable prompt patterns you would genuinely use, documented well enough that a future you — or a colleague — could pick one up and apply it correctly.

This module is longer than the others and is built around production rather than reading. It has fewer lessons and a much larger build; that is deliberate, and it is why the estimated time is higher.

This module introduces no new theory. Every requirement below is a skill from Modules 1–8, applied at the same time and written down.

Lesson 1

What Makes a Pack, Not a Pile

A pile is three prompts you happened to save. A pack is three patterns that hold to a standard.

They solve genuinely different tasks — not three variations of "summarize something".
They are documented identically, so picking one up costs no thought.
Each one says plainly when it should not be used.
Each one carries a check proportionate to what it produces.

Three is the number for a reason. One does not test whether your documentation standard holds across different work; ten becomes an exercise in volume. Three is enough to reveal whether you actually have a method. All three must be low-risk tasks: nothing involving health, money, legal matters, safety, or other people's confidential information.

Lesson 2

The Ten Fields

Every pattern in the pack is documented with the same ten fields. Each exists because an earlier module showed what goes wrong without it.

Name — what it is for, in plain words (Module 4: name it for purpose, not format).
Purpose — what problem it solves and when to reach for it (Module 4).
Intended use — the situation it was built for (Module 4).
Stable instruction — the wording that does not change (Modules 1, 2 and 4).
Variable inputs — named slots for what changes each time (Module 4).
Context requirements — what must be supplied, and what must never be baked in (Modules 3 and 7).
Constraints — the limits any acceptable result must respect (Module 2).
Expected output — the form the result should take (Modules 2 and 5).
Verification step — what you check before using it (Modules 5, 6 and 7).
Do not use when — the conditions that make it wrong (Modules 4, 7 and 8).

Fields 9 and 10 are the ones learners skip and the ones that make a pack trustworthy. A pattern without a check is a habit; a pattern without a stopping condition is a trap. Where relevant, a pattern should also carry audience and language as explicit variables — Module 8 applies to any pattern whose output reaches a reader who does not share your conventions.

Lesson 3

Reviewing Your Own Work

Before the pack is finished, run it against this checklist honestly.

Do the three patterns solve genuinely different tasks, or is one a variation of another?
Does each have a name that says its purpose?
Could someone else fill the slots correctly without asking you what they mean?
Is anything baked in that should be a variable — a date, a name, an audience, a constraint?
Is any real personal or confidential information sitting in a fixed part?
Does each verification step name what specifically to check, not just "review the output"?
Does each "do not use when" describe a real situation you can imagine happening?
Would you actually use these next week?

Throwing one out is a legitimate result. A pack of two patterns you trust is better work than three you do not, and saying so is part of the assessment.

A Pack Multiplies Whatever It Carries

A pack is reused more often than a single prompt and is more likely to be shared. Everything from Module 7 therefore applies with more force: no real personal, customer, employee, health or financial information in any fixed part of any pattern; synthetic examples only; and a verification step that names what to check.

All three capstone tasks must be low-risk by design. If a pattern you want to build touches health, money, legal matters or safety, that is the signal to choose a different task — not to add a disclaimer.

Human + AI

You integrate the system. Each earlier module handed you one responsibility; the capstone is where you decide how they fit together for work you actually do. An assistant can produce output inside any pattern you build — it cannot decide which tasks deserve a pattern, what must never be baked in, or which of your own patterns you should not trust.

There is no complete prompt system and no pack that handles everything.
Set it up once and reuse forever is the failure this whole module is built against.
A pack you would not actually use next week is not finished work.

Capstone — build your pack

≈28 min

1. Choose three recurring, low-risk tasks from different families — for example planning, communication, learning or research, organizing information, or creative drafting. Write one sentence each on why it repeats and why it is low-risk. 2. Build pattern one using all ten fields. 3. Build pattern two for a genuinely different task, and notice what you had to change about your own documentation habits. 4. Build pattern three.

5. Test each against a second, different set of inputs — the Module 4 test. Anything that had to be fought is over-fixed; make it a variable. 6. Apply the review checklist and record your honest findings, including anything you decided to discard. 7. Write one paragraph naming which earlier modules you leaned on most, and where you found the gaps in your own understanding.

You produce the capstone artifact: three documented patterns with all ten fields each, one second-inputs test per pattern with what it revealed, a completed review checklist with honest findings, and one reflection paragraph on which modules the work drew on.

The capstone is a design exercise, so no AI assistant is required or advantaged — the reasoning is design and judgment, not generation. Working entirely on paper is a full alternative, not a reduced one: build the three patterns from supplied scenarios, with a second set of inputs for each that deliberately exposes over-fixing, and review them against the checklist. A learner with an assistant may additionally run their patterns; that is enrichment, not the requirement. The artifact is yours; nothing is submitted, stored or graded.

Key takeaways

A pack is three genuinely different patterns documented to one standard — a pile is three prompts you happened to save.
Ten fields, every time: name, purpose, intended use, stable instruction, variable inputs, context requirements, constraints, expected output, verification step, do not use when.
Fields 9 and 10 are the ones people skip and the ones that make a pack trustworthy.
A pattern without a check is a habit; a pattern without a stopping condition is a trap.
Test every pattern against a second, different set of inputs — whatever you had to fight is over-fixed.
A pack multiplies whatever it carries, so keep every fixed part synthetic and low-risk by design.
Discarding a pattern you do not trust is a result, not a failure.

Practice in Prompt Lab

Optional

Want to try what you learned with real prompts? Prompt Lab is an optional practice environment, separate from this course.

Try your reusable patterns with safe sample inputs, and refine what still needs human checking.

Practice Prompting in the Real World

Opens in a new tab. Optional practice — never required for this module, the checkpoint, your progress, the Final Assessment, the certificate or Reward Points.

Before you move on

You have built something that has to hold together, reviewed it honestly, and — if the review went as it should — thrown at least one thing away.

What remains is to look back across the whole course, see the skills as one method rather than ten lessons, and demonstrate that the judgment is yours.

Review and the Final Assessment are Module 10.