ADHD Inspired This Skill. Everyone Should Be Using It

There's an open source skill literally called "i-have-adhd." The name is the joke, and it's also the whole problem summed up: if you have attention deficit disorder, an AI answer that takes four paragraphs to reach the command is punishment. But the fix it proposes isn't just for people with the diagnosis. It's for anyone reading a technical answer tired, in a hurry, or simply out of patience for the runaround.

You ask how to fix a bug. The answer comes in four paragraphs. Compliment, context, and somewhere in the middle, the command you actually needed. It ends with "Hope this helps!"

You didn't want to be let known later. You wanted the command.

The problem is the order, not the content

Every AI agent was trained to sound helpful, and helpful usually means explaining before acting. Context first, action second, courtesy at the end. That works for someone who wants to understand the whole process. It gets in the way of someone who just needs to know what to do right now.

i-have-adhd, built by ayghri, forces that order to flip. It doesn't ask the model to "be more direct," a vague request that dissolves after the first reply. It rewrites the mandatory structure: less filter between you and the next step, more focus on what actually needs to get done.

What changes in practice

The skill forces the agent into a fixed structure: action first, context only if asked for afterward. A multi-step task comes numbered, never as a running paragraph, each step with a time estimate in minutes, not "a little while." Lists never go past five items. And the response always ends with one concrete next step, never "let me know if you need anything."

A real example from the project itself: asking for help with authentication normally returns something like:

Great question! Your auth flow has a few moving pieces: the middleware, the token verification... Looking at src/auth.ts, the verifyToken function seems to be using an older API.

With the skill active, the same answer becomes:

Run npm install jsonwebtoken@latest, then edit src/auth.ts:42.
Open src/auth.ts
Replace the verifyToken function (lines 42-58) with the snippet below
Run npm test -- auth.spec.ts
Next: paste the first error line if any test fails.

How to install and activate it

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

Activate it by typing /i-have-adhd in the chat. To keep it running every session without calling it again: touch ~/.claude/.i-have-adhd-always.

codex plugin marketplace add ayghri/i-have-adhd --ref main
codex plugin add i-have-adhd@i-have-adhd

Activate it by typing $i-have-adhd. Codex can also trigger the skill on its own when it recognizes a task that would benefit from this format.

What stays the same

The skill doesn't eliminate explanation, it just gets it out of the way of action. If you want to understand the reasoning behind something, you can still ask, and the answer comes in full. What changes is the default: by default, the agent assumes you want to act, not be talked into acting first.

Next time you scroll looking for the command buried in the middle of a paragraph, the question isn't whether the explanation was good. It's whether it needed to come before the part you actually came for.