The 10 Rules of the i-have-adhd Skill: Technical Breakdown

The i-have-adhd skill enforces 10 strict communication rules—defined in skills/i-have-adhd/SKILL.md—that require leading with executable actions, capping lists at 5 items, providing concrete time estimates, and eliminating all preamble to reduce cognitive load for ADHD users.

The i-have-adhd skill is an open-source AI behavior modifier hosted in the ayghri/i-have-adhd repository that transforms standard conversational AI outputs into ADHD-optimized directives. By encoding specific constraints into the skill manifest at skills/i-have-adhd/SKILL.md, it creates a persistent session-based filter (active until the user says "stop adhd mode") that prioritizes working memory relief and immediate executability over exploratory dialogue.

The 10 Core Rules

The skill's behavior is governed by ten specific constraints documented in the manifest file. Each rule targets a specific cognitive pain point associated with ADHD, from initiation paralysis to working memory limitations.

1. Lead with the Next Action

The very first line of every response must contain a concrete, immediately executable step—such as a command, file path, or code snippet. No introductory prose, greetings, or context-setting may precede this action. According to SKILL.md lines 35-42, this rule counters initiation paralysis by removing the friction of parsing through conversational filler to find the starting point.

2. Number Multi-Step Tasks

When a solution requires sequential actions, the response must present a numbered list where each item represents a single, bounded task. As specified in SKILL.md lines 44-52, trivial steps should be combined to keep the list concise, preventing the overwhelm that comes from seeing 15 discrete micro-tasks.

3. End with One Concrete Next Action

Every explanation must conclude with a specific, doable action that can be completed in under two minutes—such as "run npm test" or "paste the error into chat." This rule, defined in SKILL.md lines 59-66, ensures the user never faces a blank screen wondering what to do next; there is always a clear exit ramp from the current interaction.

4. Suppress Tangents

If a secondary issue arises during task execution, the skill must finish the current task first before presenting the new issue. Per SKILL.md lines 66-73, tangents are treated as separate questions rather than digressions within the current thread, protecting the user's focus from context-switching costs.

5. Restate State Every Turn

Because ADHD often involves limited working memory, each response must restate the current progress explicitly (e.g., "Step X of Y done..."). Lines 75-82 of SKILL.md mandate this progress beacon so the user always knows exactly where they are in a multi-step process without holding that information in their head.

6. Give Specific Time Estimates

Vague duration phrases like "a while" or "soon" are prohibited. Every estimate must use concrete units—minutes or hours—as enforced in SKILL.md lines 84-90. This specificity allows users to budget their finite attention resources accurately.

7. Make Completed Work Visible

The skill must immediately show the result of any change (e.g., displaying a working command output or confirming a file modification). Lines 91-97 of SKILL.md require this visibility to provide immediate dopamine feedback and confirm that the action succeeded before the user moves on.

8. Matter-of-Fact Tone for Errors

Errors are reported plainly without emotive language like "Uh oh" or "Oops." The source file at lines 98-104 mandates stating the cause and fix directly, reducing the emotional charge of debugging and treating errors as neutral information rather than failures.

9. Cap Lists at 5 Items

If a procedure exceeds five steps, it must be split into categorized sub-lists (e.g., "do now" versus "later"). This constraint, found in SKILL.md lines 105-108, prevents the paralysis that occurs when presented with an unbounded workload.

10. No Preamble, Recap, or Closing Pleasantries

The response must start immediately with the answer and end as soon as the answer is complete. As defined in lines 109-116, there are no greetings ("Let me help you with that"), no summary paragraphs, and no sign-offs ("Best regards"), eliminating all friction between the question and the solution.

How to Use the i-have-adhd Skill

Activating the skill applies these ten rules persistently across the session. The following commands control the behavior.

Activating the Skill

To enter ADHD-optimized mode, type:

/i-have-adhd

The skill remains active until explicitly deactivated, applying all ten constraints to every subsequent interaction.

Example Output

When the skill is active, responses follow the strict format outlined above:

Run `npm install jsonwebtoken` then edit `src/auth.ts:42`.

1. Open `src/auth.ts`.
2. Replace the `verifyToken` function (lines 42-58) with the snippet below.
3. Run `npm test -- auth.spec.ts`.

Next: run `npm test` and paste the first failing line.

Step 3 of 3 done: tests updated. About 5 minutes total.

This output demonstrates leading with the next action (the install command), numbered steps (3 items), a concrete next action (run npm test), state restatement (Step 3 of 3 done), a specific time estimate (5 minutes), and zero preamble or closing.

Deactivating the Skill

To return to standard conversational mode, use either phrase:

stop adhd mode

or:

normal mode

The system confirms the mode change in one line and immediately resumes standard response formatting.

Summary

• The i-have-adhd skill is defined in skills/i-have-adhd/SKILL.md and applies 10 strict formatting rules for the duration of a session.
• Rule 1 mandates leading with executable actions, while Rule 10 prohibits all preamble and closings.
• Rules 2, 3, and 9 structure tasks into numbered lists capped at 5 items, always ending with a sub-two-minute next action.
• Rules 5, 6, and 7 provide working memory support through state restatement, concrete time estimates, and visible results.
• Rule 4 prevents context switching by suppressing tangents, and Rule 8 enforces neutral, actionable error reporting.

Frequently Asked Questions

How do I activate the i-have-adhd skill?

Type /i-have-adhd in the chat interface. The skill binds to the current session and reshapes all subsequent responses according to the ten rules defined in skills/i-have-adhd/SKILL.md until you explicitly exit the mode.

Can I use this skill with any AI model?

The repository includes model-specific configurations in skills/i-have-adhd/agents/openai.yaml and skills/i-have-adhd/agents/gemini.toml, allowing the skill to interface with OpenAI and Gemini models. The core rules in SKILL.md remain model-agnostic, but the agent bindings handle provider-specific API formatting.

What happens if a task naturally requires more than five steps?

Per Rule 9, the skill splits the workflow into categorized sub-lists (such as "setup now" versus "deploy later") to keep any single list under five items. If a process is inherently linear and long, the skill presents the first five steps with a clear checkpoint, then reveals subsequent steps only after the initial batch is complete.

How do I exit adhd mode?

Say stop adhd mode or normal mode. The system will confirm the deactivation in a single line and revert to standard conversational formatting, removing the ten constraints immediately.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}