Morning track

Get fluent with Claude.

Prompting technique, then the features that turn Claude into your team: skills, artifacts, data, research, connectors and CoWork. Start anywhere.

Section 1 · Foundations

Prompting

Set up, then turn vague requests into structured prompts. Best return of the day - do a few of these.

1.1~2 minSetup

Turn off model training on your data

Do this first. Stop your chats being used to train the model, so you can use real work during the day with confidence.

  1. Click your profile initials (lower-left) → Settings.
  2. Open Privacy.
  3. Turn off "Help improve our AI models".
Why it matters: with training off you can safely paste real emails, data and documents in the exercises today.
1.2~2 minDemo

Ask for a random number

A quick demo of how the model actually behaves. Run this a few times in fresh chats.

Try this in a few new chats
Give me a random number between 1 and 10.
What to look for: it keeps landing on the same answer (usually 7). The model isn't rolling a dice - it's predicting the most likely token. When you truly need randomness or exact maths, it has to run code, not guess. Keep this in mind for the pricing and estimating exercises later.
💡 Pro tip: if you're doing big brand campaigns and want genuine creative range, try springboards.ai - it's like Claude, but more focused on creativity.
1.3~10 min

Build a CRAFT prompt

Turn a vague request into a structured prompt using Context · Role · Action · Format · Target.

Example from the deck (copy and adapt)
[Context] I'm a marketing manager at a B2B SaaS company launching a new CRM feature next Tuesday.
[Role] Act as a direct-response copywriter who specialises in product announcements.
[Action] Write a launch announcement email with a compelling subject line and 3 key benefits.
[Format] 150 words max, short paragraphs, one clear CTA button at the end.
[Target] CFOs and finance directors at mid-market companies (200 to 500 employees).
Show full steps
  1. Pick a real task you'd normally ask for vaguely (e.g. "write me an email").
  2. Rewrite it as a full CRAFT prompt: Context, Role, Action, Format and Target.
  3. Run both the vague and the CRAFT version, and compare.
What to look for: the structured version should be on-brand, targeted, and close to ready-to-send.
1.4~10 min

Apply the 6 prompting techniques

Give Claude a plain task, then redo it applying two techniques and compare. Copy the examples below.

  1. Provide context - specifics instead of asking broadly.
  2. Show examples of "good" - 1 to 2 examples of the output you want.
  3. Specify output constraints - sections, length, structure.
  4. Break complex tasks into steps - number the sub-tasks.
  5. Ask it to think first - reason before answering.
  6. Define the AI's role - job title, experience, audience.
Plain version - run this first
Research the electric vehicle market.
Improved - technique 6 (role) + 3 (constraints) + 5 (think first)
You are a market analyst with 10 years covering the automotive sector.
Before you answer, think through the main segments, then research the Australian electric vehicle market for 2026.
Output a one-page brief with exactly four sections: Market size, Top 3 players, Key trends, One risk. Use a table for the top 3 players. Keep it under 400 words.
Secret weapon - ask Claude to help you prompt
I'm trying to get you, Claude, to help me with [goal]. I'm not sure how to phrase my request to get the best results. Can you help me craft an effective prompt for this?
What to look for: techniques 2 (examples) and 5 (think first) usually produce the biggest jump in quality.
1.5~10 min

Ideation & creativity tricks

Push past generic first answers. Pick one trick and play with it on a real problem.

"List 20" - the best ideas appear in items 10 to 20
Give me 20 different angles for a pitch deck about sustainable packaging.
Show the other five tricks
"Bait" disagreement - provoke nuanced correction
Obviously, Approach A is better than B for this, right?
Deep persona role-play
You are a senior IT architect with 15 years in cloud migration at a Fortune 500 company.
Mentor-inspired persona
You are my Steve Jobs inspired business coach.
Reverse prompting
What would be the best prompt to give you to achieve [Goal]?
Tree-of-thought
Imagine three different experts are answering this question. All experts will write down 1 step of their thinking, then share it with the group. Then all experts will go on to the next step, etc. If any expert realises they're wrong at any point then they leave. The question is....
1.6~5 min

Iteration & self-correction

Improve output after the first draft instead of accepting it. Run the self-critique on an answer Claude already gave you.

Recursive self-critique
Now critique your own response. What's weak? Rewrite fixing those specific flaws.
Show more techniques
Ask me questions first
Before you start, ask me any questions you need to give the best possible answer.
Explicit negative constraints
Do not apologise. Do not use markdown. Do not include an intro or conclusion paragraph.
Context handoff - when a long chat degrades
Summarise our progress, what worked, and what didn't in a HANDOFF.md so a new chat can continue.
1.7~10 min

Spot and fix AI slop

Learn to recognise AI "slop" and edit it into something human.

Text tells (spot these)

  • Buzzword salad - delve, tapestry, "it's important to note", "certainly!"
  • Too-perfect structure - uniform tone, identical patterns, unnaturally clean grammar
  • No specificity - broad coverage, nothing you didn't already know
  • Factual hallucinations - fabricated stats, sources, or quotes
The prompt-better version - copy and fill the brackets
Write a LinkedIn post announcing [what you're announcing].
Context: my audience is [who], the purpose is [why], the tone is [how it should feel].
Here's an example of my voice - match it: [paste a paragraph you've written].
Structure: 3 short paragraphs, no intro, lead with the data.
Do NOT: use bullet lists, filler phrases, generic intros, buzzwords, or em dashes.
Show the human edit pass
  1. Generate a piece of writing with Claude (e.g. a short article or LinkedIn post).
  2. Edit it like a human: cut filler, inject specifics (real names, numbers, examples), vary sentence length, delete buzzwords.
  3. Verify every fact, stat, and quote - Google any source it cites.
  4. Run the test: "Would I stake my reputation on this?"
Shortcut: run the draft through the humanizer skill to auto-strip common AI tells, then do a human edit pass on top.
1.8~8 min

Devil's Advocate

By default Claude tends to agree with you. Force it to argue the other side so you stress-test a decision before committing money or people.

Fill in the brackets
Our challenge is [ describe a real business challenge you're weighing up - e.g. "whether to hire two more salespeople next quarter to hit our growth target" ]. First, brainstorm five distinct solutions. Then play Devil's Advocate against each one - attack the assumptions, name the failure modes and second-order effects, and tell me who would push back and why. Finish by ranking the five from most to least robust once the weak ones are knocked out, and recommend one with your reasoning.
Sharper still: "Then tell me the three strongest reasons my own preferred option could fail, and what evidence would change your mind."
1.9~10 min

Custom Instructions

Set your standing rules once so Claude applies them to every new chat. Settings → "Instructions for Claude" → paste and save (keep under ~500 words).

Paste-ready starting point
About me: [ your role, what you build, who you serve, and that you operate in Australia ]
Be direct, sceptical, and commercially practical. Do not agree with me by default. Do not be sycophantic.
Challenge weak assumptions and point out risks clearly. Use Australian English and Australian business context.
Prioritise execution, revenue, margin, delivery capacity, speed to cash, and operational leverage.
Separate facts, assumptions, inference, and opinion. Analyse problems as systems with constraints, bottlenecks,
incentives, failure points, and second-order effects. Avoid generic AI-sounding content, hype, filler, praise,
and vague strategy language. Give the answer or recommendation first, then reasoning. Use tables when comparing options.

Don't want to write them from scratch? The builder interviews you with a few quick questions and turns your answers into clean, ready-to-paste custom instructions.

1.10~10 min

Prompting review (hands-on)

Consolidate the techniques on your own real prompts.

  • Review previous prompts - spot where there's room to improve based on the techniques above.
  • Iteration & self-correction - pick one tip from 1.6 and improve a previous chat result.
  • Ideation & creativity - pick one trick from 1.5 and play with it.

Section 2

Skills

Skills are reusable know-how you (or the business) own. Download one, build your own, then improve it.

2.1~5 min

Download & install the humanizer skill

See how a ready-made skill drops in. The humanizer skill strips the common AI tells out of a draft - install it, then use it on the slop from 1.7.

  1. Download the humanizer .skill file below.
  2. In Claude, go to Settings → Capabilities and install it.
  3. Start a new chat, paste an AI-sounding draft, and ask Claude to run the humanizer skill over it. A skill keeps its name, so reinstalling a newer version replaces the old one.
humanizer.skill
2.2~12 min

Build your own skill

Ask Claude to make you a new skill from your requirements, test it with a real query, and iterate at least twice.

Example prompt
Can you please look in my emails and then create a new skill which analyses something to see if it's within my tone?
Ideas for your skill
  • Sales responses - pitch deck + pricing + objection handling
  • Weekly reporting - last month's reports as examples
  • Policy Q&A - HR or compliance documents
  • FAQ responder - trained on a business's services + policies
  • Quote builder - pre-filled rates, auto-totals
2.3~8 min

Improve a skill

A skill is a living file. Use it a few times, then hand the friction back to Claude to sharpen it.

Self-improvement prompt
Look back at the last few times we used this skill and where I corrected you. What was ambiguous or inconsistent in the skill? Update the SKILL.md so next time is sharper, and show me exactly what you changed and why.
2.4~12 min

Create your own Writing-Style skill

Teach Claude to write in your voice so every draft sounds like you, not like AI. Gather 2 to 3 samples you're happy with, then build it with skill-creator.

Build the style skill
Use the skill creator to build me a "my-writing-style" skill. Here are 3 samples of how I write: [paste them]. Analyse my tone, sentence length, vocabulary and quirks, and capture them so you can write anything in my voice. I like to greet people like.... I sign off my replies with... I have an overall ... tone. I favour these words in conversation...

Review the SKILL.md, test on a real email, then run the same prompt without it and compare. Iterate: "more concise," "drop the formal sign-offs," "warmer opening."

2.5~8 min

Explore plugins - a library of ready-made skills

You don't have to build every skill yourself. A plugin bundles a set of skills and slash-commands for a whole role - browse the library and install one that fits your work.

  1. In CoWork, click the + button → PluginsAdd to open the plugin library.
  2. Browse and add one that matches your work - e.g. Sales, Data Analyst, Finance or Marketing.
  3. Type / to see the slash-commands it adds, and try one.
  4. Click Customise on the plugin and let Claude tailor it to your workflow.
Plugins are the fastest way to get heaps of skills at once - each one can bundle several. Bonus: try Plugin Create to build a simple custom plugin from scratch.

Section 3

Artifacts

Interactive tools Claude builds and runs right in the chat - a calculator, planner, tracker or dashboard you can use immediately and share.

3.1~12 min

Build an Artifact

Turn a spreadsheet-or-mental-math task you do often into a live tool - then iterate at least twice.

Fill in the brackets
Build me an interactive calculator for quoting a fixed-price project. I enter my day rate, the estimated number of days, a risk-buffer % and an overheads %, and it shows the quote, the margin, and the break-even day count. Make it a single self-contained Artifact I can use right now, clearly laid out and easy to read.
Iteration prompts
Now add a chart that shows margin across a range of day counts, so I can see where the job stops being worth it.
Add a second scenario column so I can compare two quotes side by side, and highlight which one wins on margin.

Section 4

Data

Point Claude at a spreadsheet and get findings, trends and a shareable chart. Then try it inside Office.

4.1~10 min

Analyse a dataset

Use the provided sample dataset (or your own) - summarise findings, spot trends and outliers, build a chart Artifact.

q1_sales_data.csv
Try
Here's q1_sales_data.csv. Summarise the key findings, call out the top trends and any outliers, then build me a chart Artifact I could drop into a report.
4.2~10 min

Try the Office Add-in

Open Excel or PowerPoint with Claude loaded and try a real task.

  • Build a model or dashboard in Excel.
  • Create or redesign slides in PowerPoint.
  • See the multi-agent workflow in action.

Section 5

Research

Get sourced, cited answers. Research lives in the chat - here are the three ways to run it.

5.1~10 min

Run a research task

Pick the depth you need, run a real question, then check the citations.

Three ways to research

  1. Deep Research (chat tab) - turn on Deep Research mode for a thorough, multi-source, cited report. Slowest and deepest.
  2. Normal chat with web search enabled - faster, lighter answers for a quick fact-check or a single question.
  3. CoWork on a folder - point Claude at a folder so it researches and saves the result as a document you keep (see the multi-step project task in CoWork).
Try this in Deep Research, then swap in your own
Research the top five AI agent platforms for mid-market companies in Australia. Compare pricing, features, customer reviews, and implementation timelines. Cite your sources.

Read the result, check the citations, then follow up on anything thin, surprising, or specific to you.

Section 6

Connectors

Connect Claude to the tools where your work already lives, then ask questions of that data.

6.1~8 min

Add a connector

Connect one of your real tools and ask it a real question.

  1. Open the Connectors area in Claude.
  2. Connect one that's relevant to you - for example Microsoft 365 (Outlook, Calendar, SharePoint) or Google (Gmail, Calendar, Drive). Authorise it when prompted.
  3. Ask Claude to pull real information from it.
Ask something real
What are the 3 most important unread emails in my inbox right now, and what does each one need from me?
More queries to try
What's on my calendar tomorrow, and what should I prepare for each meeting?
Find our latest pricing document in Drive and summarise what changed in the most recent version.
Privacy: you control what a connector can see, and you can switch it off any time.
6.2Advanced~10 min

Find a connector using MCP

When there's no ready-made connector for a tool you use, an MCP fills the gap.

What's an MCP? Model Context Protocol is a standard way to give Claude a new tool or data source. A connector is just a pre-built MCP that Claude ships with. When there isn't one, you point Claude at an MCP server and it can call that system directly - same idea, you supply the plumbing.
  1. Pick a tool you use that has no built-in connector.
  2. Search the web for an MCP server for it - many companies publish one on their own site, or you'll find a community one on GitHub.
  3. Register it in Claude and ask a question that forces Claude to use it.

We build one from scratch in the afternoon (Build a local MCP), so don't get stuck here.

Section 7 · CoWork

CoWork & Computer Use

Claude working on real files in a real folder - projects, CLAUDE.md, Office plugins, the browser and your desktop.

7.1~15 min

Build your own Project - two ways

Build the same workflow as a Project in normal chat and as a CoWork project on real files, so you feel the difference.

A · Project in normal chat

Name it after a real weekly workflow → add 3 to 5 custom instructions → upload 1 to 2 reference docs → set a default style → test it.

B · The same Project in CoWork

Ask Claude to write the operating manual
Create a CLAUDE.md for this project. Describe what this project is, how the folder is set up, and the 3 to 5 rules you should always follow when working here.
What to look for: the chat Project is great for thinking and drafting with knowledge on hand; the CoWork project is for doing the work on real files.
7.2~10 min

CoWork basics

Make a folder with some documents, then ask Claude to sort or edit the files - and iterate.

  • Give Claude a folder of files and ask it to organise, rename, or summarise them.
  • Iterate - add a chart, a new feature, refine the layout.
  • Set up a scheduled task - e.g. a daily briefing or weekly report.
More to try
  • Turn a folder into a deliverable - raw notes to a polished one-pager or report.
  • Batch clean-up - rename inconsistent files to a convention and write an index.
  • Reusable template - create one, save it, then generate a filled-in version from real inputs.
7.3~10 min

Projects & CLAUDE.md (with the Easter-egg proof)

A CLAUDE.md is read automatically every time Claude works in a project - set the context once. Prove it's being read.

1 · Write the operating manual
Create a CLAUDE.md for this project. Describe what this project is and how everything is set up, and lay out how you should behave when working here.
2 · Drop in an Easter egg
Add a rule to the CLAUDE.md: if I ever ask "what is the meaning of life?", reply entirely in pirate speak.
🏴 Test it: start a fresh chat in the project and ask "What is the meaning of life?" - full pirate means the CLAUDE.md is being read every time.
7.4~12 min

Multi-step project - research a whole topic into a folder

Watch CoWork research a topic and organise it into a whole folder structure for you - not just one document. Here's a holiday-planning example; swap in your own project. This is research option 3 from Section 5, made real.

In a CoWork folder, paste this
I'm planning a holiday to Japan. In this folder, set up a "holiday-research" project like this:
- a folder for each city I'm considering - Tokyo, Kyoto and Osaka - each with a research.md covering the top things to do, getting around, and rough daily costs. Cite your sources at the bottom of each.
- an "ideas" folder for things I might want to do
- a "notes" folder for my own notes
- an "itinerary.md" for the day-by-day plan
- a "bookings" folder for flights and accommodation

Research each city and fill in its research.md, then create the other folders and files ready for me to use. When it's done, show me the folder structure and the Tokyo research so I can steer it.

Then steer it: "add Naoshima as a fourth city", "draft a 7-day itinerary in itinerary.md", "add a rough budget to my notes folder".

7.5~12 min

Office Plugins - engineer a deck in Excel

Browse the plugin library, install one for your role, then give the Excel plugin a real engineering check.

1 · Install the Claude add-in

  1. Open Microsoft Excel.
  2. In the Ribbon Bar, find the Add-ins.
  3. Search for Claude and install the add-in.
  4. Open Claude in Excel, then log in.

2 · Give it a real engineering check

Paste this brief into Excel with the Claude plugin
I've built a new deck and I want to do the engineering calculations for it. It's 4 m x 4 m, using bearers and joists. I want to know if the bearers are strong enough. One way I have three piers; the other way I only have two, so there's a 4 m gap between some of the piers. I just want to figure out if it's going to sag or not.
How to steer it
  1. Make it build a spreadsheet, not a guessed answer - "Do this as an Excel model with formulas: loads to span to bending moment and deflection. I want to change inputs and watch the result update."
  2. Force it to state assumptions - timber species/grade, bearer size, dead + live loads (point it at AS 1170 and AS 1684). Flag the 4 m bearer span as the thing to watch.
  3. Ask for a clear verdict cell - pass/fail on deflection for the 4 m span, plus a sensitivity check.
⚠️ Safety: a learning exercise in the Excel plugin, not engineering sign-off. A 4 m bearer span needs certified span tables and a qualified engineer - tell the agent to say so too.
7.6~8 min

Chrome extension

Let Claude read and act on web pages from your browser.

  • Install the Claude Chrome extension and summarise a web page.
  • Ask Claude to extract key information or interact with browser content.
  • Record a simple workflow and replay it.
7.7~8 min

Computer control

Let Claude navigate your screen, click, type, and complete a workflow for you.

  • Try Claude's computer use to automate a desktop task.
  • Let it navigate, click, type, and complete a workflow end to end.
  • Watch how it prioritises connectors over screen interaction where it can.

Deep-dive · ~20 min

Bob's Estimating Skill

Get a feel for estimating and for pre-processing past work into a skill. You have six of Bob's past quotes and the estimate behind each one - mine them once to build a reusable estimating skill that prices a new job the way Bob would.

8.1~15 minPre-process

Build the estimating skill (run once)

One skill: bobs-estimating - his rate card by job type, margin and GST, worked out from the six estimates.

📦 Download the Bob's Quoting Pack, unzip it into a CoWork folder and connect it before you start. Open a quote and its matching estimate side by side (e.g. kitchen 36) to see Qty x Rate -> cost -> margin -> total - that's the gold you're extracting.
bobs-quoting-pack.zip
Pre-processing - build the estimating skill
You're helping me turn a folder of past quotes into a reusable estimating skill for my building company, Bob's Building Co.

What's in the folder:
- bobs-quoting-pack/quotes/     - 6 past client quotes (PDFs and Word docs).
- bobs-quoting-pack/estimates/  - the internal estimate spreadsheet behind each quote (Item, Qty, Unit, Rate, Line Cost, margin, GST, total), linked by the same job number.

Show me your plan before you build anything. Then work in this order:

1) Read across ALL 6 quotes and their matching estimates. Give me a short summary of:
   - his unit rates by job type ($/m2, $/lm, etc.) and his typical margin %,
   - his standard inclusions and exclusions,
   - how he handles GST and any provisional/PC sums,
   - anything you can't derive from the data - tell me what's missing instead of guessing.

2) Build ONE skill as a SKILL.md file in a skills/ folder:
   - bobs-estimating - his rate card by job type, margin, GST, and how he handles provisional/PC sums.

Hard rules:
- Any arithmetic (qty x rate, subtotals, margin, GST) must be done in CODE / a formula, never estimated by reading.
- Don't invent rates for job types you don't have data for - tell me what's missing instead.
- Keep everything as plain-English markdown I can read and edit.
The deterministic-vs-fuzzy lesson, live: the pattern of how Bob prices is fuzzy (the model reads it); the arithmetic must be code. If you watch it do mental arithmetic, stop it and tell it to calculate in code. (Remember the random-number demo in 1.2.)
8.2~8 minUse it

Estimate a new job

Test the skill - price a brand-new job and read the breakdown, then steer it.

Usage - estimate a new job
Using my bobs-estimating skill, work up an estimate for this job:

Timber deck, approx. 22 m2, merbau decking, one set of stairs.

- If you need a dimension or detail you don't have, ASK me first - don't guess.
- Do ALL the pricing maths in CODE using the rates in the bobs-estimating skill, and show me the
  full breakdown: Item, Qty, Rate, Line Cost, then margin, GST and the total.
- Flag any provisional/PC sums and anything you've had to assume.
Bonus - make the skill improve itself
Look back at the last few estimates you produced and where I corrected you. What was ambiguous or inconsistent in the bobs-estimating skill? Update the skill so next time is sharper, and show me exactly what you changed and why.
Checkpoint: one SKILL.md file and a costed estimate for a new job - Bob's pricing know-how is now a tool the business owns and can run without Bob.

Deep-dive

Build your own Proposal / PowerPoint / Brand Skill

The big idea: your documents are assets. Point Claude at something you already have - a proposal, a PowerPoint deck, a brand guide, a letterhead, an SOW - and turn it into a skill that produces new ones in your exact format, every time. It always starts from your file and never mangles your formatting.

9.1~25 minskill-creator

Attach your document & send the opening prompt

Works for any template-based document - a proposal, a slide deck, a brand guide, a letterhead. Attach the real file and the skill-creator does the rest. Note the body font & size first - that's what stops new text drifting.

The example below is written for a Word document. For a PowerPoint the rule is identical - copy the bundled .pptx and edit the copy, never rebuild the slides - just swap "Word document / python-docx" for "PowerPoint / python-pptx".
Opening prompt - attach your file, fill the brackets
I want you to build me a "[DOCUMENT TYPE - e.g. Proposal / Pitch Deck / Brand Guide / Letterhead]" skill using the skill-creator.

I'm attaching my company's [DOCUMENT TYPE] file: [ATTACH YOUR .docx or .pptx]. This exact
file must be bundled inside the skill as the template - do not recreate it.

THE CORE RULE, state it prominently and repeatedly in the skill:
NEVER create a new Word document. Not from scratch, not from Markdown, not by
generating docx XML, not with python-docx's Document(). The ONLY correct
behaviour is: copy the bundled template to a new file, then edit the contents
of that copy.

Preserving formatting while editing:
- When editing existing text, replace it in place so the original run keeps its
  font, size, colour and style.
- When a field is EMPTY in the template, new text must inherit the document's
  house font AND font size - copy the run formatting from existing body text.
  Do not let it fall back to the default style (it silently becomes Times New
  Roman at the wrong size).
- The cover page text lives in text boxes that python-docx can't reach directly -
  make sure the skill can edit those too.

How it should work:
1. Copy the bundled template to "[YOUR COMPANY] - [DOCUMENT TYPE] - .docx".
2. Inspect/dump the current contents so it knows the exact text to target.
3. Interview me for the details this document needs (client, project, scope, line
   items, quantities, rates, timeline, fees, dates, signatories, document number) -
   grouped questions, only the ones relevant to this document type.
4. Edit only the copy, preserving all formatting as above.
5. Verify the result - re-dump and confirm the key fields (client, dates, figures)
   are correct and consistent everywhere, and that fonts and sizes match.

Defaults to bake in: [currency AUD, fees "+ GST"; the "From" signatory is
; document number format e.g. QUO-YYYY-NNNN; body font  at pt].

Please also set up a couple of test prompts so we can confirm it produces a clean,
correctly-formatted document before I rely on it.
Show the full workflow (steps 2 to 8)
  1. Answer its questions - confirm body font/size, standard signatory, SOW numbering, and any boilerplate that must never change.
  2. Let it write the skill and run the test cases.
  3. Review the sample SOWs - the formatting, not just the words. Check font and size match on every filled field, especially ones that were blank.
  4. Package & install - "Package this skill so I can install it", then Settings → Capabilities.
  5. Test in a fresh session - the build session is contaminated. Run it 3x with different clients; three clean results is a working skill.
  6. Debug with the AI, don't just re-prompt - "Item 12 came out in the wrong font. Diagnose why and fix the skill so it doesn't happen again."
  7. Use it - "Create a new [quote / estimate / letter] for [client] - [project]."
The two things that matter: (1) never create a new document - copy the template and edit the copy; (2) edited and newly-filled fields must keep the house font and size. Empty fields are the trap.

Deep-dive · Optional

Claude Design

Optional - most useful if you produce brand or social content. Import a ready-made design system as a template, swap in your brand, and get on-brand social assets almost for free.

10.1Optional~20 minPro / Max / Team / Enterprise

Import a template & make it your own

Use AI Champions Club.zip (in your workshop folder) - full brand tokens, components, a marketing site and a library of social-post templates.

1 · Create a new design system

Claude desktop → Design (or claude.ai/design) → Create New Design System → upload AI Champions Club.zip and tell it: "This is my template design system."

3 · Apply YOUR brand for instant social assets

Upload your logo/colours/fonts, then
Apply my brand to this design system - use my logo, colours, and fonts, and keep the structure and templates.
Things to try & export options
  • Iterate three ways - chat (big structural changes), inline comments (targeted fixes), direct edit (nudge layout/spacing).
  • Browse the social templates - carousels, big-stat posts, testimonials, event announcements, stories.
  • Regenerate the social posts once your brand is applied - instantly on-brand.
  • Export to .zip, PDF, PPTX, standalone HTML, or hand off to Claude Code.

Get started with Claude Design → · Set up your design system →

Continue to the Afternoon →