i-have-adhd: A Skill That Makes AI Coding Agents Answer First
A skill for your coding agent to stop it from burying the answer. ADHD-friendly output.
At a glance
- What is it?
- i-have-adhd is a plugin and skill for AI coding assistants that enforces ten output rules: lead with the answer, number steps, suppress tangents, give specific time estimates, and end every response with a concrete next action. It requires no ADHD diagnosis to be useful.
- Who is it for?
- i-have-adhd is worth installing in any coding assistant whose outputs are too verbose or too hard to scan quickly. It is particularly useful in Claude Code, where the .claude-plugin/ integration exists and the README provides explicit uninstall and reinstall commands for forks.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 10 days ago.
- What is it written in?
- Mainly Python, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on September 29, 2026, and from our analysis. They are not legal advice.
DEEP OPEN-SOURCE ANALYSIS
The Problem: AI Agents That Bury the Answer
AI coding assistants often respond with a long preamble before reaching the actionable content. A model asked why an authentication token fails might open with a summary of how JWT works, enumerate possible causes across three paragraphs, and close with "Hope this helps!" before naming the actual line to change. For engineers who need to act quickly and return to focus, that structure imposes a scanning cost on every response. The i-have-adhd project formalizes the alternative: put the next action first, number any sequence of steps, end with one concrete next instruction, and suppress tangents entirely. The README contrasts the two styles directly with a before and after example. The before version opens with "Great question!" and ends with "Hope this helps!"; the after version opens with the exact npm command to run and names the file and line to edit.
The Ten Output Rules the Skill Enforces
The full rules live in SKILL.md inside the skills/i-have-adhd/ directory of the repository. The README lists them in summary form. Rule 1: lead with the next action. Rule 2: number multi-step tasks. Rule 3: end with one concrete next step. Rule 4: suppress tangents. Rule 5: restate state every turn so the reader does not have to track context mentally. Rule 6: give specific time estimates in minutes, not vague phrases. Rule 7: make wins visible when a step completes. Rule 8: treat errors matter-of-factly, without escalating. Rule 9: cap lists at 5 items. Rule 10: no preamble, no recap, no closers. Together these rules produce responses that a reader can act on without reading the full text first. The rules are not enforced by the model at a technical level; they are instructions the model receives and follows as part of its context.
Installing i-have-adhd in Claude Code and Other Agents
The README gives an install method that works in Claude Code, Codex, Cursor, and other agents: paste the following prompt into the CLI:
Install the i-have-adhd skill/plugin from https://github.com/ayghri/i-have-adhd, refer to the repo's AGENTS.md for instructions.The agent reads AGENTS.md in the repository and follows the install steps described there. The repository top-level directory includes .agents/, .claude-plugin/, .codex-plugin/, .cursor/, .opencode/, and gemini-extension.json, indicating per-agent integration directories. For detailed install steps for a specific agent, INSTALL.md in the repository is the documented reference. The package.json shows version 0.3.0 and lists the skill under the pi.skills key, pointing to the skills/ directory.
Forking and Customizing the Rules
The README describes a four-step process for customizing the skill by forking the repository, editing SKILL.md, and reinstalling from the fork. The full command sequence from the README is:
claude plugin uninstall i-have-adhd
claude plugin marketplace remove i-have-adhd
claude plugin marketplace add <your-username>/i-have-adhd
claude plugin install i-have-adhd@i-have-adhdThe first two commands remove the upstream copy. The third registers the fork under the same name from the marketplace. The fourth installs it. Restart the coding assistant and reinvoke the skill after completing these steps. This pattern allows teams to add, remove, or reword rules without maintaining a divergent branch indefinitely; the SKILL.md file is the only file that needs to change for most customizations.
What i-have-adhd Does Not Change
The skill is a set of instructions added to the model's context. It does not modify the underlying model, change its training, or enforce output length at a technical level. A sufficiently complex question may still produce long output if the rules allow the content the response genuinely requires. The skill also does not apply when the user invokes a different skill that overrides the output format. The evals/ directory in the repository suggests the authors maintain test cases for the skill's output behavior, but the README does not describe what those tests cover or what pass rate is expected. The skill cannot guarantee consistent behavior across model updates, since the model's interpretation of the rules may shift across versions.
One Alternative: Writing Custom Agent Instructions Without a Skill
The same output behavior can be approximated by adding system-level instructions directly to a coding assistant's configuration. Claude Code, for instance, allows custom instructions in CLAUDE.md or via the --system-prompt flag. The difference between that approach and i-have-adhd is maintenance: the skill bundles the rules into a named, versioned, shareable package that can be installed, forked, and updated via the plugin marketplace. A manual CLAUDE.md entry is per-project and per-engineer. Teams working across multiple repositories or multiple engineers benefit from the shared skill approach over per-project system prompt additions.
Maintenance Status
The last push to the main branch was on 2026-09-19, and the current package version is 0.3.0. The repository has no GitHub releases; versioning is tracked in package.json. The project is MIT-licensed and credits "The Adult ADHD Tool Kit" by J. Russell Ramsay and Anthony L. Rostain as an adaptation source, with the README noting the adaptation was made for how an LLM should respond, not how a human should organize their day. The repository is multilingual: README translations exist for Simplified Chinese, Spanish, Portuguese (Brazil), Japanese, Vietnamese, Korean, Farsi, Thai, and Arabic, each in .github/readme/.
Editorial conclusion
i-have-adhd is worth installing in any coding assistant whose outputs are too verbose or too hard to scan quickly. It is particularly useful in Claude Code, where the .claude-plugin/ integration exists and the README provides explicit uninstall and reinstall commands for forks. The skill does not change the underlying model's capabilities; it reshapes how responses are presented. Teams who prefer detailed explanatory output from their assistant will want to evaluate whether the rules are too constraining before adopting it. Customize by editing skills/i-have-adhd/SKILL.md in a fork and reinstalling from that fork.
Frequently asked questions
What is the i-have-adhd skill for AI agents?
i-have-adhd is a plugin for AI coding assistants that enforces ten output rules: lead with the next action, number multi-step tasks, suppress tangents, restate state each turn, give specific time estimates, and end with one concrete next step. It works with Claude Code, Codex, Cursor, and other agents.
Do I need an ADHD diagnosis to use the i-have-adhd skill?
No. The README states explicitly: "No ADHD diagnosis needed." The skill is designed for anyone who finds AI agent output too verbose or too slow to scan.
How do I customize the i-have-adhd skill rules?
Fork the repository, edit skills/i-have-adhd/SKILL.md with your changes, then uninstall the upstream copy and reinstall from your fork using the four claude plugin commands shown in the README.
Community notes