Categories of Rules in I-Have-ADHD: The 10 Core Principles and Implementation Guide

The I-Have-ADHD skill organizes its assistant behavior into 10 concrete output rules, supported by persistence settings, cognitive fact bases, explicit exception clauses, and pre-send validation checks—all defined in the canonical SKILL.md file.

The ayghri/i-have-adhd repository provides a specialized skill that reshapes how AI assistants generate responses for readers with ADHD. Understanding the categories of rules in I-Have-ADHD is essential for developers who want to implement structured, action-oriented output that accommodates working-memory limitations and dopamine-scarce cognition. These rules are codified in skills/i-have-adhd/SKILL.md and function as an instruction set for LLM behavior modification.

Session Persistence and Cognitive Foundations

The skill begins with two meta-categories that establish context before the actionable rules apply.

Persistence Settings

According to skills/i-have-adhd/SKILL.md (lines 15-20), Persistence defines rules that remain active for the entire session until explicitly disabled by the user. This ensures consistent behavioral modification across multiple interactions without requiring re-activation.

What ADHD Changes About Reading

Lines 23-30 of SKILL.md establish five cognitive facts that motivate all subsequent rules:

• Working-memory limits
• Friction between knowing and doing
• Difficulty starting tasks
• Uniform time perception
• Scarce dopamine availability

These facts serve as the psychological foundation for the rigid output structure that follows.

The 10 Core Output Rules

The heart of the categories of rules in I-Have-ADHD consists of ten specific behavioral mandates governing assistant output.

Rule 1 – Lead with the Next Action

As specified in lines 33-40 of SKILL.md, the first line of any response must contain an immediate, doable action. This eliminates preliminary explanations and places the user directly into execution mode.

Rule 2 – Number Multi-Step Tasks

Defined in lines 42-50, this rule requires breaking work into concise numbered lists where each step is clearly bounded. This accommodates working-memory limitations by making progress explicit and resumable.

Rule 3 – End with One Concrete Next Action

Lines 57-63 mandate that every response closes with a single, easy-to-execute next step, reducing activation energy for the subsequent interaction.

Rule 4 – Suppress Tangents

Lines 64-70 establish that the assistant must finish one issue before introducing another, preventing the cognitive fragmentation that occurs when multiple threads compete for attention.

Rule 5 – Restate State Every Turn

According to lines 73-80, the assistant must remind the reader of the current step and progress at every turn, compensating for the "out of sight, out of mind" characteristic of ADHD working memory.

Rule 6 – Give Specific Time Estimates

Lines 82-88 require concrete timing (minutes/hours) instead of vague qualifiers like "soon" or "a while," addressing the uniform time perception common in ADHD.

Rule 7 – Make Completed Work Visible

Specified in lines 89-95, this rule requires explicitly showing what now works rather than burying accomplishments in recaps, providing necessary dopamine feedback.

Rule 8 – Matter-of-Fact Tone for Errors

Lines 96-101 dictate that errors be presented with cause and fix stated without emotive language, preventing shame spirals that can derail task completion.

Rule 9 – Cap Lists at 5 Items

Lines 103-106 establish that longer lists must be split into "do now" versus "later" or "must" versus "nice-to-have" categories, respecting the limited cognitive bandwidth for task initiation.

Rule 10 – No Preamble, Recap, or Closing Pleasantries

Lines 107-113 of SKILL.md require the removal of generic openings and closings, keeping the entire response focused exclusively on actionable content.

Exception Handling and Quality Assurance

The rule categories include explicit mechanisms for deviation and validation.

When to Break the Rules

Lines 117-126 of SKILL.md define specific situations where the default rules are overridden, including:

• Detailed explanations when requested
• Safety confirmations
• Breaking out of debug spirals
• User-initiated conversational tangents

These exceptions prevent rigid adherence from becoming counterproductive.

Pre-Send Validation Check

Lines 128-138 establish a final sanity-check requiring verification that the first and last lines of any output convey the necessary actions, ensuring Rule 1 and Rule 3 compliance before the response reaches the user.

Implementation Examples

The following patterns demonstrate the rule categories in practice.

Invoking the skill activation:

User: /i-have-adhd
Assistant: Run `npm install jsonwebtoken`, then edit `src/auth.ts:42`.

Multi-step task execution following Rule 2:

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 the tests and paste the first failing line.

Applying Rule 10 (no preamble):

Login now works with magic links. Try: `npm run dev`, then open `/login`.

Implementing Rule 9 (cap at 5 items):

Do now:
- Install dependencies
- Run the linter
- Fix failing tests
Later:
- Update documentation
- Add CI workflow

Key Repository Files

Beyond skills/i-have-adhd/SKILL.md, the implementation relies on:

• .opencode/command/i-have-adhd.md – Runtime description for OpenCode integration
• hooks/hooks.json – Declares the skill as a cross-runtime hook
• README.md – High-level usage overview

These files collectively enable the skill's deployment across different AI assistant platforms.

Summary

The categories of rules in I-Have-ADHD provide a comprehensive framework for ADHD-optimized AI interaction:

• Persistence maintains session-wide behavioral context
• Cognitive facts ground the rules in ADHD psychology
• 10 core rules structure output for working-memory limitations and dopamine management
• Exception clauses allow flexibility when rigid adherence would harm outcomes
• Pre-send validation ensures output quality through automated checking

Frequently Asked Questions

What is the I-Have-ADHD skill?

The I-Have-ADHD skill is a behavioral instruction set defined in skills/i-have-adhd/SKILL.md that modifies how AI assistants generate responses. It imposes structure—such as leading with actions, numbering steps, and removing pleasantries—to accommodate the cognitive patterns associated with ADHD, specifically working-memory constraints and task initiation difficulties.

How do I activate these rules in a session?

According to the persistence settings in lines 15-20 of SKILL.md, the rules activate when a user invokes the skill (typically via a command like /i-have-adhd) and remain in effect for the entire session until explicitly disabled. The hooks/hooks.json file enables this persistence across compatible runtimes.

When should the AI break the I-Have-ADHD rules?

Lines 117-126 of SKILL.md specify four primary exception scenarios: when the user requests detailed explanations, when safety confirmations are required, when breaking out of debug spirals, and when the user deliberately initiates conversational tangents. These exceptions ensure the rules serve the user rather than forcing rigid compliance in inappropriate contexts.

What is the pre-send check?

The pre-send check is a quality assurance step defined in lines 128-138 of SKILL.md. Before delivering any response, the system verifies that the first line presents an immediate action (per Rule 1) and the last line presents a concrete next step (per Rule 3), ensuring the output meets the skill's structural requirements.

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