How to use the i-have-adhd agent skill: setup and examples

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md file with instructions that an AI agent loads when it is relevant. The i-have-adhd skill is a style skill: it shapes the format and tone of replies rather than adding new tools.

Once active, the agent applies rules such as: start with the first action to take, write a numbered list for multi-step work, end with one concrete next action, and avoid side tangents by offering them as a separate question at the end. It also states that errors should be described plainly as cause and fix.

• Leads with the action, not context or a plan
• Turns multi-step work into short numbered steps
• Restates current state every turn (for example, step 3 of 5 done)
• Gives specific time estimates such as about 15 minutes
• Highlights finished work in concrete terms
• Keeps the style on until you say "stop adhd mode"

How do you install the i-have-adhd skill?

Download the ZIP file from this page and unzip it. For a personal install, place the unzipped folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

In the Claude apps, you can instead upload the skill under Settings → Capabilities → Skills.

• Personal install: ~/.claude/skills/i-have-adhd/
• Project install: .claude/skills/i-have-adhd/ inside your project
• Claude apps: Settings → Capabilities → Skills, then upload

How do you turn on the style?

Invoke the skill by typing /i-have-adhd. The style then applies to every reply for the rest of the session, not only the next one. The skill is set so that the model does not invoke it on its own, so you start it deliberately.

To turn it off, say "stop adhd mode" or "normal mode". The agent confirms in one line and returns to its default style.

What requests work well with the skill?

The skill works with ordinary requests. Ask for the task you want done, and the replies will take the shape described above. For coding work, ask for the change you need and the agent will give the file, the lines to edit, and the command to check it.

If you are stuck on a multi-step job, say so and ask where you are. The agent restates progress each turn so you do not have to hold the whole plan in your head.

• Ask for a single fix and get the file, line and next command first
• Ask for a multi-step setup and get a numbered list with one action per step
• Ask what is left and get the current step plus one next action

What should you know before relying on it?

The skill trims content. If you want long, detailed explanations, it is the wrong style: it can hide context you need. You can say "explain" or "walk me through" to get a fuller answer for that request.

Inside an agent tool that enforces its own output format, the tool's rules win over this skill. The skill's own text says the harness outranks it.

The skill has no listed requirements, runs no scripts and uses no network access, according to its listing. It is a set of written instructions.

• Not suited to readers who want long explanations
• Yields to output formats required by other tools
• No scripts run, no network use, nothing listed as required