If you use AI for coding every day, you kinda know the drill.
You feed the agent a failing test, a traceback, or a bug. You just want the terminal command to run or the exact line number to fix.
Instead, you get a wall of text:
“Great question! Let me think about this. Your auth flow has a few moving pieces… One approach would be… By the way, you might also want to look at…”
This happens a lot with my ML projects.
Sometimes, you have to scroll past three paragraphs of conversational filler just to find the one npm install command you actually needed.
It breaks your flow.
It drains your focus.
Enter ayghri/i-have-adhd.
Despite the name, you don’t need an ADHD diagnosis to benefit from it.
It’s an open-source GitHub repository that provides a custom plugin for Claude Code, designed to ruthlessly cut the fluff and format AI outputs for pure execution.
What It Actually Does
This plugin acts as a strict guardrail that forces Claude to change how it communicates.
It stops the AI from acting like a chatty pair programmer and turns it into a high-efficiency terminal tool.
Here is the difference in action:
The Before:
A sprawling response explaining the history of an API, suggesting a refactor, offering unsolicited advice about dependency versions, and signing off with a cheerful “Hope this helps!”
The After:
Run npm install jsonwebtoken@latest, then edit src/auth.ts:42.
1. Open src/auth.ts
2. Replace verifyToken (lines 42–58) with the snippet below
3. Run npm test -- auth.spec.ts
Next: paste the first failing line if any test fails.The Engine: 10 Core Rules
The magic of this plugin lies in its SKILL.md file.
It imposes 10 strict rules on Claude's output.
But here are the ones that make the biggest impact on a developer's daily workflow:
- Lead with the next action. No preamble. No “Here is what I found.” Just the immediate next step.
- Number multi-step tasks. Cap lists at 5 items. If it takes more than 5 steps, the AI needs to break it down to prevent cognitive overload.
- Specific time estimates. Instead of “this will take a bit,” it must estimate in actual minutes.
- Make wins visible. Keeps momentum going when you’re deep in a grueling debugging session.
- No closers. Absolutely no “Hope this helps!”, “Let me know what you think,” or recap summaries.
How to Set It Up
If you’re already using Claude Code locally, adding this to your environment takes just a few seconds.
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-adhdOnce installed, just type /i-have-adhd in your Claude Code session to initialize it.
If you ever need the AI to explain a complex architecture concept and want the conversational tone back,
simply use /plugin disable i-have-adhd.
Countless developers use LLMs as a tool to figure out what they’re doing.
But interacting with AI can actually cause context switching.
When you have to parse 400 words of text to extract a single variable name, you’re doing unnecessary cognitive work.
By constraining the AI to output only actionable, numbered steps, ayghri/i-have-adhd preserves your mental cache.
It proves that sometimes the best way to improve an AI agent isn't by giving it more context:
it's by forcing it to speak less.
In case we are meeting for the first time, come over here, it’ll be worth the roller coaster of articles that are gonna come up in the next few weeks.