i-have-adhd Agent Skill Guide

Key Takeaways#

• i-have-adhd is a coding-agent skill/plugin for shorter, action-first answers.
• The project is not medical software; it is an output-formatting skill for developer agents.
• The README documents ten rules, including numbered steps, one concrete next action, visible wins, and no filler closers.
• The repository includes integration markers for Claude Code, Codex, Cursor, OpenCode, Gemini CLI, Qwen, Kimi, and related agent/plugin systems.

What this resource covers#

i-have-adhd is useful when an AI coding assistant keeps burying the answer under preamble, caveats, and generic closing text. The repository describes it as a skill that stops a coding agent from burying the answer and makes the output ADHD-friendly. No ADHD diagnosis is required. The practical target is simple: make an assistant lead with the next action, keep steps numbered, suppress tangents, and end with one specific next step.

That framing matters for developers because agent productivity is often lost in response shape, not only model quality. A capable model can still waste time if it starts every answer with context recaps, offers too many alternatives, or hides the command you need halfway down the page. This skill gives the agent behavioral rules that make each turn easier to scan and easier to execute.

What the skill changes#

The README contrasts a verbose answer with a direct one. In the improved version, the assistant starts with the command to run, lists numbered edits, tells the user which test to execute, and ends with one concrete next step. The documented rules include leading with the next action, numbering multi-step tasks, making wins visible, giving matter-of-fact errors, ranking long lists, and avoiding preamble, recaps, and closers.

For everyday coding work, that means the agent should say what to do before explaining why. If a task has three steps, it should number them. If a test fails, it should ask for the first failing line instead of starting a broad debugging essay. If a long list is necessary, it should group and rank it instead of dumping a wall of options.

Where it fits#

Use i-have-adhd as a lightweight agent-output preference for teams that already use tools such as Claude Code, Codex, Cursor, OpenCode, Gemini CLI, Qwen, or Kimi. It is especially helpful for bug fixing, command-heavy setup tasks, and review loops where the next action matters more than a long explanation.

The repo is also a good template for teams building their own agent skills. The rules are small enough to audit, fork, and adapt. The README explains that users can customize the skill by forking the repository, editing skills/i-have-adhd/SKILL.md, and installing the fork into their agent workflow.

Practical adoption checklist#

1. Read the README and INSTALL.md in the official repository.
2. Install the plugin or copy the skill into your agent environment.
3. Run it on a real coding task, not a toy prompt.
4. Check whether the agent now gives commands, file paths, and next steps earlier.
5. Fork the rules if your team needs different limits for detail, explanation, or tone.

Limitations#

This is a prompt/skill layer, not a correctness guarantee. It changes how the assistant communicates; it does not make the model more accurate, run tests automatically, or replace human review. It also should not be treated as medical advice or accessibility software. Its value is narrower and practical: reducing answer clutter for coding-agent workflows.

Source#

Official repository: https://github.com/ayghri/i-have-adhd