Fit your own exoskeleton.
Eleven stages. About thirty minutes. Coffee in hand. The cleanest path from blank laptop to a working AI assistant that thinks alongside you.
Yesterday you didn't know you could do this. Today you'll have a suit that thinks alongside you. Tomorrow you'll wonder how you ever did without one.
Begin Stage 01 →You're cleared for stage 01.
A working stack — yours, on your laptop, with you tomorrow.
Same laptop. Before, and after.
Meet your six senses.
Most AI tools have one sense: read what you type, reply with what they know. Your suit gets six — each a specialized expert it can consult. Together they make the AI roughly 10× sharper on real work.
A note before we start.
Eleven stages. Each one fits a single plate of armor. By the end you're wearing the whole suit.
You can't break anything. Every command is reversible. The worst case is a window closes and we open it again. Pause anywhere — the page remembers where you are.
Your kit.
- A laptop. Mac, Windows, or Linux — all three work. Windows needs one extra setup step we'll handle in Stage 02.
- Coffee, tea, water. Something to sip between stages.
- ~30 minutes. Pause anywhere — progress saves automatically.
- Willingness to type into a window. Like a recipe. You don't need to understand every ingredient.
At the end you'll have an AI helper that knows your laptop, your project, and you — and that will still be there tomorrow.
Install Claude Code.
The helmet is the suit's brain — eyes and ears that see your work. Without it, nothing else has anywhere to live.
Claude Code is the version of Claude that runs on your laptop and can read + write real files. (The "Code" in the name just means "files." Don't let it scare you.)
Get Claude Code Or get Codex CLI (OpenAI)Click whichever AI you prefer — the exoskeleton skill works with both. (Anthropic and OpenAI agreed on the same skill format in late 2025, so the bundle is dual-target.) The page detects your OS and gives you the right installer. Run it like any normal app install.
Open the terminal.
The visor is the window the suit looks through — your text-based interface to it. You'll spend about five minutes total in here for the whole guide.
On macOS: press Cmd+Space, type Terminal, press Enter.
On Windows: press the Start button, type WSL, press Enter. If WSL isn't installed yet, install it first:
Install WSL2 (Microsoft docs)Or in PowerShell, type:
Reboot when it finishes. Then find WSL in your Start menu.
On Linux: open the app called Terminal from your applications menu. (On most distros, Ctrl+Alt+T also opens one.)
Make a folder for your project.
The spinal port is where the suit anchors to one specific task. Every project gets its own port. Or skip this entirely — in Stage 05 the AI will make the folder for you and download the suit itself.
If you want to make the folder yourself:
Press Enter after each line. Change my-project to whatever name you want.
~/Documents is the Linux home Documents folder — not your Windows Documents. They're separate. It works fine inside WSL.Three plates in. You've come further than most ever do. The hard plates are next, but you won't feel them — Claude wears the weight.
Wake Claude up in the folder.
Until now your AI knows about the world. This command tells it to pay attention to your world — this folder, your project, you.
If you installed Claude Code:
If you installed Codex CLI instead:
Either way, press Enter.
One prompt. The AI does the rest.
No more typing for you. Paste the prompt below and Claude (or Codex) will fetch the suit, install it, restart itself if needed, and walk you through the rest — all on its own.
Paste it into the AI prompt. Press Enter.
That's it. From here, the AI handles every technical thing:
- ⚡ Picks the right skills folder for your AI
- ⚡
git clones the exoskeleton bundle for you - ⚡ Reads its own SKILL.md to know what to do next
- ⚡ Switches into Concierge Mode automatically
- ⚡ Asks you four simple questions (next stage)
git clone, you'll see a few lines about "Cloning into '.claude/skills/exoskeleton'..." ending in done. Then it confirms it's switching to concierge mode and asks the first calibration question.github.com/c-merkel/exoskeleton. Most common reasons a clone fails: git isn't installed yet on your machine (your AI will spot this and walk you through installing it), no internet, or a typo in the URL. Still stuck? Email hello@christianmerkel.com.Claude asks. You answer.
Calibration is how the suit learns your shape. Four easy questions. Plain English. You'll never need to know the technical names of the things Claude is configuring.
"What's your project called?"
Just the name. "My Storefront" or "Mom's Bakery". It can change later.
"What does it do?"
One sentence. Plain English. "A website to book my services and remember every client."
"What database do you want?"
Say MariaDB. Free, mature, fast — the canonical default for this stack.
"What programming language?"
Say "you pick — something modern and easy." Claude will choose well.
Six plates in. The longest stage is next, but you're mostly watching. Sip slowly. Claude does the heavy work.
Watch the suit assemble.
Claude does the heavy lifting now. You'll watch a series of green checkmarks accumulate. This is the stage where your laptop becomes a workshop.
The big moves Claude makes:
- ⚡ Install Docker — boxes that keep your programs from arguing with each other. Claude links you to Docker Desktop's download page when needed.
- ⚡ Set up your GitHub connection — work backed up to the cloud, signed with your identity.
- ⚡ Wire all six senses — Cartographer, Archivist, Keeper, Translator, Witness, Visitor. Each gets installed and verified.
- ⚡ Build the workshop — your project's actual website with a blank starting page you can visit right away.
A few times Claude will ask you to:
→ Click a link in your browser to authorize something.
→ Type your laptop password.
→ Wait for a green ✓ before continuing.
Copy the unstick prompt
Step into your new workshop.
First time the suit moves. Claude will point you to a link — that's your project running on your laptop, your private rehearsal stage.
This link works only on the laptop running your workshop — not the public internet.
Speak to your suit.
Until now Claude was setting itself up. Now you ask for something real. The suit responds — that's the moment everything changes.
Go back to your terminal. Try one of these (click to copy):
- "Add a welcome page that says my business name."
- "Make me a form where someone can leave their name and email."
- "What can you help me with on this project?"
Press Enter. Watch.
Nine plates in. The hard work is behind you. Two more — small ones — and the suit is yours forever.
Save your work.
The memory bank is the suit's checkpoint system. Save state every time you finish something — and you'll never lose a day's work.
Claude can save a snapshot any time:
Do this often — once a session, end of day, after a big change.
Close it. Come back tomorrow.
The final plate is the rhythm itself. The suit will be exactly where you left it whenever you put it on.
Close the terminal. Close the browser tab. Shut your laptop. It's all saved.
Tomorrow, or whenever:
2.
cd ~/Documents/my-project3.
claude4. Just start talking. Claude remembers everything.
That's the rhythm. Open. Talk. Save. Close.
Three first missions for the suit.
Pick one. Paste the prompt to Claude. See what happens. None takes more than 20 minutes — and each one teaches you something different.
When something feels off — three magic prompts.
You won't always know what to type. Expected. Normal. Three prompts that always get you unstuck. Click any to copy. Bookmark this page — you may want them tomorrow.
One more thing: Claude is never offended. Not a person, doesn't get tired. Ask the same question seven ways if you need to. The thing bends to you.
You did this.
An hour ago you'd never opened a terminal. Right now you wear an AI exoskeleton on a real project on a real laptop. That's not nothing. That's everything.