Brooke Wright
Brooke Wright · @wright_mode
FREE SETUP GUIDE

Make Claude talk to your ADHD brain

Get the free setup that makes Claude answer in short, do-this-now steps instead of a wall of text you have to re-read three times.

Copy-paste ready No code required Works in any AI chat 5-minute setup
Something went wrong. Please try again.

No spam. Unsubscribe anytime.

The 3-step setup that makes Claude answer in short, do-this-now steps Jump to the setup → Join the Membership
Wright Mode · Free Resource

The ADHD-Friendly Claude Setup

Three tiny tweaks that make Claude answer in short, action-first steps, so you actually do the thing instead of drowning in another wall of text.

Copy-paste ready No code required Works in any AI chat 5-minute setup

What's inside

Here's the whole thing. Pick the route that matches how you use Claude. You don't need all three, and Step 2 works on its own if you never want to install anything.

1

The easy install (Claude Code)

If you use Claude Code (the terminal one), just tell Claude to install it for you. Paste this in and it'll do the rest:

Install the i-have-adhd skill from github.com/ayghri/i-have-adhd and confirm it's active.

Prefer to run it yourself? This one line does the same job. It adds the skill and installs it. After it finishes, type /i-have-adhd in your chat to switch it on:

claude plugin marketplace add ayghri/i-have-adhd && claude plugin install i-have-adhd@i-have-adhd

On Claude.ai or Cowork instead?

There's no one-click install for the web app or Cowork yet. Two options: in Cowork, download the SKILL.md file from the repo (github.com/ayghri/i-have-adhd) and add it as a custom skill. On Claude.ai, just use the paste-in prompt in Step 2. It does the exact same job in any chat, no install needed.

2

The no-install version (paste this into any chat)

This is the one to grab if the word "install" makes your eye twitch. Paste it at the start of any Claude chat and it'll answer your ADHD brain from then on. It's the 10 rules from the skill, reworded so a human can read them:

For the rest of this chat, answer me like this: 1. Start with the action. Your first line is something I can actually DO, not background and not a plan. 2. Number anything that takes more than one step. One action per step. 3. Finish with a single next step I can do in under two minutes. 4. Park your tangents. If a second issue exists, deal with the first, then raise the second as its own question at the end. 5. Remind me where we are ("step 3 of 5") every time, because I won't remember between messages. 6. Give time estimates in real units. "About 10 minutes," not "a bit." 7. Tell me plainly what now works. Don't bury the win inside a recap. 8. If something breaks, just say what caused it and how to fix it. Skip the "oops." 9. Cap every list at 5 things. If it's longer, split it into "do now" and "later." 10. No preamble, no recap, no sign-off pleasantries. Start with the answer. Keep this up until I say "stop ADHD mode." (Rules adapted from the open-source i-have-adhd skill by @ayghri, github.com/ayghri/i-have-adhd, MIT licensed.)
3

Make it stick

So you're not re-pasting it forever. If you installed the skill in Claude Code, run touch ~/.claude/.i-have-adhd-always once and it stays on across every session. If you're using the paste-in version, drop those 10 rules into your Claude Project instructions (or a saved note you paste at the top of new chats). Set once, forget about it.


It's not you. The default AI answer is built to look thorough: long, hedged, wrapped in "great question!" and a polite recap at the end. For an ADHD brain that's a nightmare. The one thing you actually need is buried in paragraph three, and you pay the tax of reading the whole thing twice to find it.

You ask a simple question. You get an essay. Somewhere in it is the answer, but by the time you've hunted it down you've lost the thread of what you were even doing. That's the tax. This setup just removes it. Same smart AI, none of the wading.

😮‍💨 Before

"Great question! There are a few things worth considering here. First, it's important to understand the context... [280 more words] ... and so, taking all of that into account, you'll want to change the value on line 14. Let me know if you'd like me to explain any of this further!"

😮‍💨 After

Change line 14 to true.

That's the fix. Next step: save the file and refresh. Takes about 20 seconds.

Same information. One of them you can act on before you get distracted. That's the whole point.


Once the setup's in, these little phrases make it even calmer to work with. Steal them.

👉

Ask for the next single step only

"Just give me the next one step" stops Claude dumping the whole roadmap when your brain only has room for one thing right now.

Demand dot points

"Answer in dot points, not paragraphs." Prose hides the action; a list can't. Works on any topic, not just tech.

One question at a time

"Interview me one question at a time." Turns an overwhelming form into a chat you can actually finish.

🅿️

Park the tangent

"Hold that thought till the end." Your idea doesn't vanish. Claude parks it and hands it back once the main thing's done.

Cap the list

"Give me 5 max." A list of 12 options is a decision you'll never make. Five is a choice you can actually make today.

🔁

Ask it to recap where you are

Coming back after a break? "Remind me what we were doing and what's next." No shame in it. That's what it's for.


The install and the 10 rules come from a brilliant open-source project called i-have-adhd, built by a developer who goes by @ayghri. It's blown up to 7.7k stars on GitHub, because it turns out a LOT of us want our AI to just get to the point.

It's MIT licensed, which means it's free to use and share. I've reworded the rules into plain English for the paste-in version, but full credit goes to the original. If you found this useful, go star the repo. That's how good free tools stay alive.

⭐ The original skill

i-have-adhd by @ayghri · MIT licensed · 7.7k stars

github.com/ayghri/i-have-adhd →


Ready to go deeper?

Take your skills to the next level with these resources.