10 Response-Style Rules Enforced by the 'i-have-adhd' Skill

The 'i-have-adhd' skill enforces 10 rules that make AI responses concise, action-oriented, and ADHD-friendly: lead with immediate actions, number multi-step tasks, end with one concrete next step, suppress tangents, restate state every turn, give specific time estimates, make completed work visible, use matter-of-fact error tones, cap lists at five items, and eliminate preambles or closing pleasantries.

The i-have-adhd repository by Ayoub Ghriss (ayghri) provides a Cursor skill designed to reformat AI assistant responses for readers with ADHD. According to the source code in skills/i-have-adhd/SKILL.md, the skill intercepts all output and applies ten strict formatting rules that reduce cognitive load and keep users focused on executable next steps.

---

What Is the 'i-have-adhd' Skill?

The skill is a behavioral modifier that runs inside Cursor IDE environments. When active, it transforms every assistant response to match a concise, action-oriented output style defined in the canonical skill file. The repository maintains two synchronized copies of the definition: the canonical skills/i-have-adhd/SKILL.md and a runtime mirror at .cursor/skills/i-have-adhd/SKILL.md.

---

All 10 Response-Style Rules Explained

Each rule below is extracted directly from the source file with line references preserved.

Rule 1: Lead with the Next Action

The first line of every response must be a concrete step the reader can execute immediately (SKILL.md#L33-L41).

No context-setting, no greetings—just the action. This prevents the reader from scanning through filler text to find what matters.

Rule 2: Number Multi-Step Tasks

When a solution requires more than one step, present a numbered list where each item is a single bounded action (SKILL.md#L42-L55).

Each number creates a clear stopping point and sense of progress.

Rule 3: End with One Concrete Next Action

After any explanation, close with exactly one short action the reader can do in under two minutes (SKILL.md#L57-L63).

This prevents decision paralysis by narrowing infinite possibilities to one.

Rule 4: Suppress Tangents

Finish the current issue before introducing a new one; offer the second issue as a separate question (SKILL.md#L64-L70).

The skill explicitly blocks "by the way" diversions that fragment attention.

Rule 5: Restate State Every Turn

Remind the reader of the current step number and what has been completed (SKILL.md#L73-L80).

Working memory support: each response re-anchors the user in context.

Rule 6: Give Specific Time Estimates

Use concrete units (e.g., "≈ 15 minutes") instead of vague phrases (SKILL.md#L82-L88).

Time-blindness accommodation: specific numbers enable realistic planning.

Rule 7: Make Completed Work Visible

Explicitly show what now works, using concrete commands or URLs (SKILL.md#L89-L95).

Immediate positive reinforcement prevents the "wait, did that work?" uncertainty loop.

Rule 8: Matter-of-Fact Tone for Errors

State the cause and fix without informal lamentations (SKILL.md#L96-L101).

No "Oops!" or "Sorry about that!"—just the facts and the solution path.

Rule 9: Cap Lists at Five Items

If a list exceeds five items, split it into "do now" vs. "later" or "must" vs. "nice-to-have" (SKILL.md#L103-L106).

The five-item limit matches typical working memory capacity; forced prioritization reduces overwhelm.

Rule 10: No Preamble, Recap, or Closing Pleasantries

Begin with the answer and finish when the answer is complete; avoid filler phrases (SKILL.md#L107-L114).

Every word must earn its place. No "In summary," no "I hope this helps," no "Best regards."

---

Example Response in 'i-have-adhd' Style

This snippet from the source analysis demonstrates all rules in practice:

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

# Output after the fix

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

Next: run `npm test` and paste the first failing line. (≈ 2 minutes)

• Rule 1: Immediate action ("Open src/auth.ts")
• Rule 2: Numbered, bounded steps
• Rule 5: Implicit state through sequential order
• Rule 7: Visible result ("Login now works...")
• Rule 3: Single next action with Rule 6 time estimate
• Rule 10: No preamble or sign-off

---

Key Source Files in the Repository

File	Purpose
skills/i-have-adhd/SKILL.md	Canonical definition of all 10 rules (source)
.cursor/skills/i-have-adhd/SKILL.md	Runtime mirror for Cursor IDE (source)
README.md	Repository overview and installation
AGENTS.md	Runtime entry points and skill loading logic

The two SKILL.md files must stay synchronized; the Cursor runtime loads from .cursor/ while the canonical definition lives in skills/.

---

Summary

• The 'i-have-adhd' skill enforces 10 strict response-style rules designed for ADHD cognition
• Every rule targets a specific friction point: working memory limits, time blindness, decision paralysis, or distraction susceptibility
• Source authority: skills/i-have-adhd/SKILL.md lines 33-114 contain the complete rule definitions
• Implementation: Cursor IDE loads the skill from .cursor/skills/i-have-adhd/SKILL.md at runtime
• Result: AI responses become scannable, actionable, and cognitively affordable

---

Frequently Asked Questions

How do I enable the 'i-have-adhd' skill in Cursor?

Copy the skill files to your Cursor configuration directory as specified in README.md. The skill activates automatically for all assistant responses once the .cursor/skills/i-have-adhd/SKILL.md file is present and referenced in your agent configuration.

Can I modify the 10 rules for my own needs?

Yes—edit skills/i-have-adhd/SKILL.md directly. The repository treats this file as the source of truth; maintainers recommend syncing changes to the .cursor/ mirror manually until automated synchronization is implemented.

Why cap lists at exactly five items?

The five-item limit corresponds to Miller's Law regarding working memory capacity. According to the source code at SKILL.md#L103-L106, exceeding this threshold without explicit prioritization increases cognitive load and reduces completion rates.

Does the skill work with other AI assistants besides Cursor?

The repository is currently configured for Cursor IDE. The AGENTS.md file maps runtime entry points specific to Cursor's agent system; porting to other platforms would require adapter logic to intercept and reformat responses according to the same ten rules.

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