THE FUTURE MARKETACADEMY
Your classroomRoom 01 · AI Lab
School of AI

Prompt Systems: production-grade prompt libraries

Step 7 of 7 in AI Lab

Next in your path: 1.11.4 Claude Practitioner: design, teach and audit Claude workflows

What you'll be able to do
  • Turn scattered prompts into a team prompt library with owners, standards and a simple catalogue.
  • Write every library prompt to one standard: a header, clear variables, labelled parts, examples and an output spec.
  • Version prompts with a change log, statuses and a way to roll back.
  • Add guardrails: untrusted-content rules, refusal wording and limits on what the AI may do.
  • Attach a test suite to each prompt, and re-run it when the model, the app or the prompt changes.
  • Review and audit other people's prompts, and teach colleagues to write and maintain them.

Before this course: Prompt Testing: evaluate and improve your prompts (course 1.2.3). You already know how to write specific, measurable success criteria, build a test set with edge cases and a held-out group, score answers with a rubric, compare versions fairly with one change at a time, find error types and keep a change log. Earlier, course 1.2.2 taught chains, roles and tags, structured outputs, examples and self-checks, and course 1.2 taught the basics and a simple prompt library. This course scales one tested prompt into a maintained system that a team can trust. No coding is required. Where Anthropic's pages show code or Console tools, we describe the same ideas so you can do them by hand in documents and spreadsheets.

Words in this course 53
AgentAI (artificial intelligence)APIArticleChange controlChannelContext windowCSVData analyticsData minimisationDistributionElectric currentEval (evaluation)File pathFillHeld-out test setHTTPSHuman in the loopLarge language model (LLM)Least privilegeLetLibrary (package)LoopMeanMixture and solutionPathPeer reviewPlain formProjectProject (Claude)Project instructionsPromptPrompt injectionPrompt libraryPrompt templateProofingRed-teamingRiskRubricSample
Open the Vocabulary Hall
01

From a pile of prompts to a prompt library

In simple words

A kitchen where every cook keeps their own recipes in their pocket is messy. Nobody knows which recipe is the newest or which one burns the soup. A prompt library is a shared recipe book: one place, clear names, someone responsible for each recipe, and notes on what changed.

Key facts
  • A prompt library here is a shared, organised collection of prompts that a team uses on purpose, each with an owner, a status and a test record. A folder of copied chats is not a library.
  • Course 1.2 suggested a simple library of your best prompts with blanks. A prompt system goes further: standards, versions, guardrails, tests and review, so the library stays reliable when people and models change.
  • Anthropic's engineering article on building effective agents advises finding the simplest solution possible and adding complexity only when needed. Apply that here: start with a few high-value prompts, not hundreds.
  • What belongs in a library: prompts used repeatedly, by more than one person, where wrong output has a cost. What does not belong: one-off chats and anything that includes private data in the prompt text itself.
  • A catalogue is a simple table (a spreadsheet is enough) with one row per prompt. This layout is a practical suggestion:
ColumnExample
ID and nameEM-03 Customer email triage
Purpose (one sentence)Sort customer emails and draft replies for a person to approve
OwnerAna
Version and statusv1.2, approved (draft, tested, approved, retired)
Risk levelMedium: sent to customers after review
Tested on (model or app, date)Date and the Claude app or model name used
LocationLink to the prompt note
  • Every prompt has one owner who is responsible for fixing and testing it. A prompt with no owner decays.
  • Where the library lives matters. Use a place the team already uses (a shared document or drive), with access that matches your organisation's rules. Do not put confidential data into prompts or examples unless your employer's policy allows it.
  • Claude Projects hold custom instructions and uploaded files for a workspace, and Anthropic's help pages say sharing a project with teammates is for Team and Enterprise plans. Plan details change; check the app before you rely on this.
Step by step
  1. Collect every prompt your team really uses. Ask each person to send their top three.
  2. Remove duplicates and one-off prompts. Keep the ones used at least weekly or where a mistake is costly.
  3. Create the catalogue table with the seven columns and add one row per prompt.
  4. Give each prompt an ID, a one-sentence purpose and one owner.
  5. Mark each prompt's current status honestly (most will be "draft" because none is tested yet).
  6. Put the library in a shared place and set who may edit and who may only read.
  7. Check: someone new could find the prompt they need in under one minute and see who owns it.
Watch on YouTube
The prompting playbookClaude · YouTube · youtube.com

Videos by other creators on YouTube, for reference. They are not made by or affiliated with The Future Market.

LESSON 01From a pile of prompts to a prompt library
LESSON 01 / ANIMATED LESSON / AI VOICE00:00 / 01:00
Read the video script
[Host] "A pile of prompts in pockets is not a system."
[Screen: spreadsheet with ID, purpose, owner, version, status, risk] "One table: what each prompt is for, who owns it, which version, and how risky it is."
[Screen: statuses: draft, tested, approved, retired] "Every prompt has a status. Most start as draft."
[Text on screen: Start small] "Start with a few prompts that matter. Add more only when you need to."
Try it now

Build a catalogue of at least five real prompts from your team with owners, risk levels and statuses.

ClaudeStuck? Copy a ready-made prompt for Claude

Copy this, open Claude, paste it, and Claude will walk you through the lesson step by step.

For a grown-up: Claude is for adults (18+). Copy this, paste it into Claude, and use the answer to explain the lesson to your child.

I'm learning "Prompt Systems: production-grade prompt libraries" on The Future Market. Lesson 1: "From a pile of prompts to a prompt library".

What the lesson teaches:
- A prompt library here is a shared, organised collection of prompts that a team uses on purpose, each with an owner, a status and a test record. A folder of copied chats is not a library.
- Course 1.2 suggested a simple library of your best prompts with blanks. A prompt system goes further: standards, versions, guardrails, tests and review, so the library stays reliable when people and models change.
- Anthropic's engineering article on building effective agents advises finding the simplest solution possible and adding complexity only when needed. Apply that here: start with a few high-value prompts, not hundreds.
- What belongs in a library: prompts used repeatedly, by more than one person, where wrong output has a cost. What does not belong: one-off chats and anything that includes private data in the prompt text itself.
- A catalogue is a simple table (a spreadsheet is enough) with one row per prompt. This layout is a practical suggestion:
- Column - Example
- ID and name - EM-03 Customer email triage

The exercise I'm working on:
Build a catalogue of at least five real prompts from your team with owners, risk levels and statuses.

I'm stuck. Please:
1. Explain the main idea in plain, simple words with one everyday example.
2. Walk me through the exercise one small step at a time. Ask me what I've done so far before giving the next step.
3. Don't just give me the final answer. Help me do it myself, and check my work when I share it.
I'm a parent helping my child (about 8 years old) with a lesson called "From a pile of prompts to a prompt library" from the course "Prompt Systems: production-grade prompt libraries".

Here's what the lesson says:
A kitchen where every cook keeps their own recipes in their pocket is messy. Nobody knows which recipe is the newest or which one burns the soup. A prompt library is a shared recipe book: one place, clear names, someone responsible for each recipe, and notes on what changed.

The activity:
Build a catalogue of at least five real prompts from your team with owners, risk levels and statuses.

Please explain this so an 8-year-old understands: short sentences, a fun everyday example, and no hard words.
Then give me 3 simple questions I can ask my child to check they understood, and a way to do the activity together, one tiny step at a time.
Open Claude
Sources
My notes
02

One standard for every prompt

This lesson is part of full access: every course and every level for one payment.

03

Versions, status and change control

This lesson is part of full access: every course and every level for one payment.

04

Guardrails

This lesson is part of full access: every course and every level for one payment.

05

A test suite for every prompt

This lesson is part of full access: every course and every level for one payment.

06

Review and audit prompts

This lesson is part of full access: every course and every level for one payment.

07

Teach and run the system

This lesson is part of full access: every course and every level for one payment.

++

Final project

This lesson is part of full access: every course and every level for one payment.

??

Quiz

  1. What makes a prompt library different from a folder of copied chats?

  2. What does Anthropic's advice on complexity say?

  3. Why use named variables such as {{customer_email}} in a library prompt?

  4. Name the four statuses used in this course.

  5. What should you do instead of editing an approved prompt in place?

  6. Why record the model or app a version was tested on?

  7. What is the difference between a jailbreak and an indirect prompt injection?

  8. What does Anthropic say about leak-proofing a prompt?

  9. What is a must-still-pass group?

  10. When should you re-run a prompt's test suite?

  11. What is Anthropic's "key test" for a prompt?

  12. What does a quarterly audit of the library look for?

EX

Final exam

This lesson is part of full access: every course and every level for one payment.

CE

Certificate

This lesson is part of full access: every course and every level for one payment.