What the i-have-adhd skill does
The skill starts from five facts about reading with ADHD: working memory is small, knowing is not doing, starting is the hardest step, vague time estimates all feel the same, and progress has to be visible. Its rules follow from those:
- Lead with the next action. The first line is a command, a path or a step, not context.
- Number multi-step work, with the fewest steps that still work.
- End with one concrete next action that takes under two minutes.
- Suppress tangents. A second issue becomes a separate question at the end.
- Restate state every turn, such as "Step 3 of 5 done".
- Give specific time estimates in minutes or hours.
- Make completed work visible, cap visible lists at about five items, use a matter-of-fact tone for errors, and drop preambles, recaps and closing pleasantries.
It also says when to break the rules: when you ask for an explanation, before a destructive action, after three "still broken" turns, or when the request is genuinely ambiguous.
When it runs
Type /i-have-adhd. The skill turns off automatic invocation, so Claude will not switch it on by itself. Once on, it applies to every reply in the session until you say "stop adhd mode" or "normal mode".
Good to know
In our test it did not change the technical plan, only its shape. It is released under the MIT license and also ships settings for Gemini and OpenAI agents. See more coding skills.

