The 10 Canonical Response Rules for the i-have-adhd Skill: A Complete Technical Guide

The i-have-adhd skill enforces 10 strict output rules—codified in skills/i-have-adhd/SKILL.md—that require every response to lead with an immediate action, use numbered steps for multi-step tasks, suppress tangents, and eliminate conversational fluff.

The i-have-adhd skill in the ayghri/i-have-adhd repository defines a rigid IO canonical response contract designed to make AI outputs instantly actionable for readers with ADHD. These ten rules govern everything from sentence structure to list length, ensuring that every interaction remains focused, concrete, and free of cognitive overhead.

The 10 Canonical Response Rules Explained

Each rule is enforced for the entire session unless the user explicitly disables them (e.g., by saying “stop adhd mode”). The canonical source resides at skills/i-have-adhd/SKILL.md.

1. Lead with the Next Action

The first line of every response must be a concrete step the reader can perform immediately. According to the source code at lines 33-41, the output cannot open with context, greetings, or explanations; it must jump straight to the executable instruction.

2. Number Multi-Step Tasks

When a solution requires several steps, present them as a numbered list where each step represents a single bounded action. As implemented in lines 42-55, this format prevents the reader from parsing dense paragraphs to identify discrete actions.

3. End with One Concrete Next Action

After any explanation, the response must finish by naming one thing the reader can do in under two minutes. Lines 57-63 specify that this final line acts as a “release valve,” ensuring the user never faces a vague conclusion.

4. Suppress Tangents

Finish the current issue before introducing a new one. Lines 64-70 mandate that related but separate issues must be offered as distinct questions rather than merged into the current response, preventing context switching.

5. Restate State Every Turn

Re-announce the current step and overall progress on each turn. Lines 73-79 require this so the reader does not need to hold prior context in working memory; the response is self-contained.

6. Give Specific Time Estimates

Provide concrete time windows (e.g., “≈15 minutes”) rather than vague descriptors like “soon” or “quickly.” Lines 82-88 emphasize temporal specificity to aid time-blindness management.

7. Make Completed Work Visible

Explicitly state what now works and how to verify it. Lines 89-95 require the response to summarize the new state and provide a verification command or check, closing the feedback loop.

8. Matter-of-Fact Tone for Errors

Describe failures without emotive language; give cause and fix directly. Lines 96-101 prohibit apologetic or alarming phrasing, instead requiring a neutral statement of the error, its cause, and the remediation step.

9. Cap Lists at Five Items

If a list grows beyond five items, split it (e.g., “do now” vs. “later”). Lines 103-106 enforce this to prevent overwhelm; the reader receives only the immediately relevant subset.

10. No Preamble or Closing Pleasantries

Start straight with the answer and finish when the answer ends. Lines 107-114 ban generic greetings, sign-offs, or “hope that helps” closures, reducing word count and social overhead.

Implementation Examples

The following snippets demonstrate correctly-shaped responses according to the 10 canonical rules.

Example 1: Simple One-Step Fix

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

Rule compliance: The first line is a concrete action (Rule 1). There is no preamble or sign-off (Rule 10).

Example 2: Multi-Step Task

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 copy the first failing line.

Rule compliance: Steps are numbered (Rule 2). The final line gives a single actionable next step under two minutes (Rule 3). A time estimate could be added, e.g., “≈5 minutes” (Rule 6).

Example 3: Handling an Error

Test fails at `auth.spec.ts:42`: expected 200, got 401.
Cause: missing `Authorization` header.
Fix: add `Authorization: Bearer ${token}` to the request.
Run the test again.

Rule compliance: Uses a matter-of-fact tone (Rule 8) and ends with a concrete next action (Rule 3).

Example 4: List Longer Than Five Items

Do now:
1. Update dependencies.
2. Fix lint errors.
3. Write missing unit tests.

Later:
4. Refactor the authentication module.
5. Improve documentation.

Rule compliance: Splits a long list into “do now” vs. “later” (Rule 9), keeping each segment under the five-item cap.

Source Files and Repository Structure

The 10 canonical response rules are defined, exposed, and synchronized across three primary files in the ayghri/i-have-adhd repository:

• skills/i-have-adhd/SKILL.md: The canonical source of truth containing the ten response rules and their enforcement logic (lines 33-114).
• AGENTS.md: Provides the repository-wide mapping of agents and points to the skill’s location for runtime discovery.
• .cursor/skills/i-have-adhd/SKILL.md: A mirror of the canonical skill for Cursor-based runtimes; kept in strict sync with the source file to ensure consistent behavior across editors.

Summary

The i-have-adhd skill transforms standard AI outputs into executive-function-friendly formats by enforcing strict structural constraints:

• Every response must lead and end with a single, concrete action (Rules 1 and 3).
• Numbered lists organize multi-step work, capped at five items per list (Rules 2 and 9).
• State restatement and specific time estimates reduce working memory load (Rules 5 and 6).
• Matter-of-fact error handling and zero conversational fluff keep cognitive overhead minimal (Rules 8 and 10).
• All rules reside in skills/i-have-adhd/SKILL.md and propagate to .cursor/skills/i-have-adhd/SKILL.md.

Frequently Asked Questions

Where are the 10 canonical response rules for the i-have-adhd skill defined?

The rules are defined in skills/i-have-adhd/SKILL.md at lines 33-114. This file serves as the single source of truth; a mirrored copy exists at .cursor/skills/i-have-adhd/SKILL.md for Cursor IDE integration.

How do I temporarily disable the i-have-adhd skill rules during a session?

Users can disable the skill for the current session by explicitly requesting cancellation, such as saying “stop adhd mode.” Once disabled, the IO canonical response contract no longer applies until the session resets.

Why is the list length capped at five items?

Rule 9 caps lists at five items to prevent decision paralysis and overwhelm. According to lines 103-106 of SKILL.md, any list exceeding five items must be split into themed sub-lists (e.g., “do now” vs. “later”), ensuring the reader processes only immediately actionable information.

Do the canonical rules allow any exceptions?

Yes. While the default flow enforces all ten rules, the skill defines explicit exceptions for scenarios requiring safety-critical confirmations or when the user explicitly requests an explanatory, non-actionable response. These exceptions override the standard “no preamble” and “lead with action” constraints temporarily.

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