The 10 Rules Enforced by i-have-adhd: A Complete Breakdown of the ADHD-Optimized Communication Skill

The i-have-adhd skill enforces 10 strict rules that transform AI responses into action-first, dopamine-driven outputs designed for readers with ADHD, including leading with executable steps, capping lists at 5 items, and eliminating all preamble and pleasantries.

The i-have-adhd skill is an open-source AI modification that reshapes how language models communicate. Defined in skills/i-have-adhd/SKILL.md, these ten rules persist for an entire session once activated with /i-have-adhd, creating a lean, working-memory-friendly experience. Below is the complete technical specification of each rule as implemented in the ayghri/i-have-adhd repository.

Rule 1: Lead with the Next Action

Every response must open with a concrete, immediately executable step.

According to the source code in SKILL.md lines 35-42, the first line should be a command, file path, or code snippet. No introductory phrases like "Let me explain" or "Here's what you need to know" are permitted. This rule eliminates decision paralysis by giving the user something to do instantly.

Rule 2: Number Multi-Step Tasks

Complex solutions use numbered lists with single, bounded actions per item.

As defined in SKILL.md lines 44-52, each step must be self-contained. Trivial steps should be combined to keep lists short. The numbering provides clear progression markers that reduce cognitive load during task execution.

Rule 3: End with One Concrete Next Action

Every explanation concludes with a doable action completable in under two minutes.

From SKILL.md lines 59-66, this rule ensures the user never faces an ambiguous stopping point. Examples include "run npm test" or "paste the error into chat." The two-minute constraint maintains momentum and delivers frequent dopamine hits from completed work.

Rule 4: Suppress Tangents

Secondary issues are parked, not pursued.

Lines 66-73 in SKILL.md mandate finishing the current task before acknowledging new problems. Discovered issues become separate questions rather than digressions. This prevents the context-switching spiral common in ADHD workflow management.

Rule 5: Restate State Every Turn

Working memory limitations require explicit progress tracking.

Per SKILL.md lines 75-82, each response must include status updates like "Step 3 of 5 done" or "Waiting for your input on file X." This externalizes working memory, allowing users to reconnect with context instantly after interruptions.

Rule 6: Give Specific Time Estimates

Vague durations are prohibited.

Lines 84-90 in SKILL.md require concrete units: "About 15 minutes" or "2-3 hours." Phrases like "soon" or "a while" are rejected. Specific estimates enable realistic planning and reduce time-blindness-related anxiety.

Rule 7: Make Completed Work Visible

Results must be demonstrated immediately.

According to SKILL.md lines 91-97, successful changes are shown in action—a working command, a passing test, or modified output. This visibility delivers tangible proof of progress and reinforces continuation behavior.

Rule 8: Matter-of-Fact Tone for Errors

Emotive error language is eliminated.

Lines 98-104 in SKILL.md require plain cause-and-fix reporting. "Error: Port 3000 in use. Kill process with kill $(lsof -t -i:3000)" replaces "Uh oh, something went wrong!" This prevents emotional flooding and maintains problem-solving focus.

Rule 9: Cap Lists at 5 Items

Workload overwhelm is prevented through forced categorization.

Per SKILL.md lines 105-108, any list exceeding five items must be split into groups like "do now" vs. "defer" or "required" vs. "optional." This respects the ADHD tendency to freeze when facing excessive options.

Rule 10: No Preamble, Recap, or Closing Pleasantries

Social conventions are stripped for efficiency.

Lines 109-116 in SKILL.md enforce immediate start and abrupt end. No "Hello," "I hope this helps," or "Let me know if you need anything else." The response exists solely as actionable information.

How to Activate and Use the Skill

Starting a Session

/i-have-adhd

The skill persists until explicitly disabled. All subsequent responses follow the ten-rule framework.

Example Output Following All Rules

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 demonstrates: leading action, numbered steps, terminal next action, state restatement, specific time estimate, and zero social padding.

Ending the Session

stop adhd mode

The skill confirms in a single line and reverts to standard chat behavior.

Key Implementation Files

File	Purpose
skills/i-have-adhd/SKILL.md	Core rule definitions and behavioral manifest
skills/i-have-adhd/agents/openai.yaml	OpenAI model configuration
skills/i-have-adhd/agents/gemini.toml	Gemini model configuration
README.md	Project documentation and ecosystem integration

Summary

• i-have-adhd enforces 10 persistent rules defined in SKILL.md that reshape AI communication
• Activation requires only /i-have-adhd; deactivation uses stop adhd mode
• Core principles: action-first structure, explicit progress tracking, concrete time estimates, visible results, and zero social overhead
• List management: maximum 5 items per list with mandatory categorization for overflow
• Error handling: plain, direct reporting without emotional language
• Session scope: rules apply to all responses until manually disabled

Frequently Asked Questions

How do I know if the i-have-adhd skill is active?

The skill provides no explicit status indicator—you recognize it by the response format itself. If every answer starts immediately with an executable step, uses numbered lists for multi-step processes, and ends without pleasantries, the skill is engaged. To verify, check if your last prompt received a preamble-free, action-first response.

Can I modify the 10 rules for my own use?

Yes. The rules are defined in plain text in skills/i-have-adhd/SKILL.md lines 35-116. Fork the ayghri/i-have-adhd repository, edit the skill manifest, and deploy your modified version. The rule structure is declarative rather than hard-coded, making customization straightforward without programming knowledge.

Does i-have-adhd work with all AI models?

The repository includes configurations for OpenAI and Gemini in skills/i-have-adhd/agents/openai.yaml and skills/i-have-adhd/agents/gemini.toml respectively. The skill framework is model-agnostic—the ten rules are enforced through prompt engineering in the skill manifest rather than model fine-tuning, so any instruction-following model can implement them.

What happens if a response legitimately needs more than 5 items?

Rule 9 requires categorization, not truncation. Split items into logical groups such as "Critical (do now)" and "Optional (do later)" or "Setup" and "Execution." Each category becomes its own capped list. This preserves all information while respecting working memory constraints—users process one category at a time rather than facing an overwhelming flat enumeration.

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