Free skill · MITClaude CodeAction-first1,100+ ★

If you have ADHD, stop using Claude like this.

Claude often knows exactly how to fix your problem — but first it makes you read 14 paragraphs to find it. You ask one question and get a whole essay before it tells you what to actually do. This free skill, "I Have ADHD", forces Claude to give you the action first: do this, then this, then this.

Get the repo free See the rules

You commented ADHD — here's the repo, the one-time install, and the trick to make it always-on so you never type the command. No diagnosis required; it's just good when your brain is already overwhelmed.

What changes

Same answer. Minus the essay.

It doesn't make Claude dumber — it reshapes the output so an overwhelmed brain can act on it. The information is identical. The order isn't.

Before

"Great question! Let me think about this. Your auth flow has a few moving pieces: the middleware, the token verification, and the cookie handling. Looking at src/auth.ts, the verifyToken function (around lines 42-58) seems to be using an older API… Hope this helps! Let me know if you want to dig deeper."

After

Run npm install jsonwebtoken@latest, then edit src/auth.ts:42.
1. Open src/auth.ts
2. Replace verifyToken (lines 42–58)
3. Run npm test -- auth.spec.ts

Next: paste the first failing line if any test fails.

10rules, all action-first

1line to install, one to run

MITfree and open-source

1,100+GitHub stars

Setup · about 2 minutes

Install once, then one command.

It's a Claude Code plugin with one skill inside. Clone it, register it, install it. Then type the command whenever you want action-first mode — or make it permanent in the next section.

Install the plugin

Three commands in your terminal

git clone https://github.com/ayghri/i-have-adhd ./i-have-adhd
claude plugin marketplace add ./i-have-adhd
claude plugin install i-have-adhd@i-have-adhd

# then, inside Claude Code:
/i-have-adhd

Verify with claude plugin list — you want to see i-have-adhd (enabled). If /i-have-adhd isn't in autocomplete, restart Claude Code; the plugin index loads at startup.

Turn it on

Type /i-have-adhd and it flips

From that point in the session, Claude leads with the action, numbers the steps, and drops the "Hope this helps!" It applies to everything — coding, debugging, planning, even casual questions. Turn it off anytime with /plugin disable i-have-adhd.

Sauce · make it permanent

Never type the command again

The best move: make it always-on. Drop this into your ~/.claude/CLAUDE.md (global, every project) and Claude follows the rules from message one — no /i-have-adhd needed:

## Output style

Always follow the rules in the `i-have-adhd` skill:
action-first, numbered steps, no preamble, no closers,
state restated each turn.

iPut it in a project's CLAUDE.md instead if you only want action-first mode on certain repos.

Get the repo on GitHub Read the full SKILL.md ↗

Under the hood

The 10 rules it enforces.

This is the whole engine. Read them and you'll see why the output feels different — and you can tune any of them by editing SKILL.md.

1Lead with the next action.
2Number multi-step tasks.
3End with one concrete next step.
4Suppress tangents — finish one thing first.
5Restate state every turn (“step 3 of 5 done”).
6Specific time estimates — minutes, not “a bit”.
7Make wins visible, not buried in a recap.
8Matter-of-fact errors: cause, then fix.
9Cap lists at 5 items, ranked.
10No preamble, no recap, no closing pleasantries.

Extra sauce

Three upgrades most people miss.

Not in Claude Code?

A paste-anywhere version

The plugin is Claude-Code-only. But the idea is portable — paste this at the top of any chat (claude.ai, ChatGPT, Gemini) to get the same action-first behavior for that conversation:

Answer me action-first. Rules:
- First line is the thing to do (a command, path, or step) — not context.
- Number any multi-step work. One action per step.
- No preamble ("Great question", "Let me…") and no closers ("Hope this helps").
- Cap lists at 5 items. If longer, split into "do now" vs "later".
- End with ONE concrete next step I can do in under 2 minutes.
- If I say "explain", go long — but still no preamble, no closer.

It knows when to back off

It won't leave you in the dark

Action-first doesn't mean it hides the "why". The skill overrides itself when it should: say "explain" or "walk me through" and it goes long (still no preamble). Before anything destructive — rm -rf, force push, a migration — it stops and confirms. And if you're three turns deep in "still broken", it stops guessing and asks one diagnostic question instead.

Make it yours

Tune the rules to your brain

Every rule lives in plain English in skills/i-have-adhd/SKILL.md. Want harder caps, different time-estimate wording, or a rule about emojis? Edit the file and re-invoke /i-have-adhd. It's your output style, not a black box.

Straight answers

The honest FAQ.

Do I need an ADHD diagnosis?

No. The repo literally says “no ADHD diagnosis needed.” If Claude's walls of text overwhelm you — and they overwhelm most people — this helps.

Does it make Claude dumber?

No. It changes the shape of the answer, not the content. Same fix, same reasoning — just the action first and the fluff gone. Ask it to explain and it'll go as deep as you want.

Does it only work for coding?

No — the skill triggers on any message, including planning and casual chat. Anywhere Claude would normally preamble, it gets to the point instead.

Is it free?

Yes. MIT-licensed, open-source at github.com/ayghri/i-have-adhd. No paid tier.

Credit where it's due

I didn't build this — it's by Ayoub G. (ayghri), MIT-licensed, and loosely based on The Adult ADHD Tool Kit by Ramsay and Rostain — adapted for how an LLM should respond. Go star it. If it saves you a scroll, tag me on Instagram — I repost the good ones — or bring your tweaks to the weekly call on Skool.

Keep going

More ways to make Claude actually usable.

Skills and setups that fix Claude's defaults — the way it writes, the way it spends your limits, the way it designs. Free, and they stack.

Systems & Agents

5 Claude skills that fix its terrible writing

Claude writes generic content by default. Five free Claude Code skills fix it one failure mode at a time: dumbify (8th-grade readi…

Read guide

Systems & Agents

Stop burning your Claude limits — the Fable + Sonnet setup

You hit Claude limits because your most expensive model does everything, including the grunt work. Three built-in Claude Code sett…

Read guide

Systems & Agents

Kimi K3 inside Claude Code harness is a cheat code

Kimi K3 just dropped — a frontier-level model at roughly a fifth of the price. The play isn't to replace Claude; it's to make Kimi…

Read guide

Systems & Agents

42 Claude design skills that kill AI slop

AI-generated design all looks the same — purple gradient, Inter, four cards in a grid. These 42 SKILL.md skills for Claude Code, C…

Read guide

Browse the full library

Keep going

Fix Claude's defaults, one skill at a time.

One email a week with the free skills and setups that make AI actually usable — the installs, the tweaks, and the exact steps.

Join the newsletter Browse all resources