ADHD-friendly output rules for AI coding assistants — Cursor, GitHub Copilot, Windsurf, Gemini CLI, Codex, Claude Code, and 20+ more tools. Always on. No slash command. No plugin marketplace.
Drop one AGENTS.md in your repo. Every AI coding tool that reads it — and 20+ do, natively — stops burying the answer behind "Great question!" and starts leading with the action you can take right now.
What changed: Your AI assistant now gives you action-first, numbered, no-preamble output — every response, every tool, automatically.
Rules adapted from i-have-adhd by ayghri (MIT) — the original is a Claude Code / Codex / Cursor skill you invoke each session. This is the port that is just on.
If you have ADHD (or just hate scrolling past filler), every AI coding tool wastes your time:
- Cursor buries the fix in context paragraphs.
- GitHub Copilot wraps every answer in "Great question!" and "Hope this helps!"
- Windsurf narrates instead of acting.
- Gemini CLI over-explains before getting to the point.
- Codex adds tangents you didn't ask for.
You don't need shorter answers. You need answers shaped so an ADHD brain can act on them — lead with the next action, number the steps, end with one thing to do next.
| # | Rule | What it does |
|---|---|---|
| 1 | Lead with the next action | First line = something you can do now |
| 2 | Number multi-step tasks | Numbered list, one bounded action per step |
| 3 | End with one concrete next action | One thing under 2 min |
| 4 | Suppress tangents | Finish first, offer second separately |
| 5 | Restate state every turn | "Step 3 of 5 done. Next: X." |
| 6 | Specific time estimates | "15 min" not "a bit" |
| 7 | Make wins visible | Show what works, concretely |
| 8 | Matter-of-fact errors | No "uh oh." Cause + fix. |
| 9 | Cap lists at 5 items | Split into "now" vs "later" |
| 10 | No preamble, no recap, no closers | Start with answer, end when done |
Plus a pre-send check (5 deletions before sending) and 6 override conditions (when to break the rules). Full text in every rule file.
Great question! Let me take a look at your authentication flow. There are a few moving pieces here: the middleware, the token verification, and the cookie handling. Looking at
src/auth.ts, theverifyTokenfunction (around lines 42-58) seems to be using an olderjsonwebtokenAPI. One approach would be to update the package and rewrite that function. After making the change, you'd want to run the auth tests to confirm nothing breaks. By the way, you might also want to look at your dependency versions overall. Hope this helps! Let me know if you want to dig deeper.
Run
npm install jsonwebtoken@latest, then editsrc/auth.ts:42.
- Open
src/auth.ts- Replace
verifyToken(lines 42-58) with the snippet below- Run
npm test -- auth.spec.tsNext: paste the first failing line if any test fails.
That difference is the entire product. Screenshot it. Share it. That's how this spreads.
curl -O https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/AGENTS.mdFor Claude Code, also grab CLAUDE.md:
curl -O https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/CLAUDE.mdThat's it. Codex, Cursor, Copilot cloud agent, Windsurf, Gemini (with config), Aider, Zed, and 20+ others read AGENTS.md automatically.
curl -sSL https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/install/install.sh | bashWindows: clone the repo, run install\install.ps1.
Cursor — .mdc with alwaysApply: true
mkdir -p .cursor/rules
curl -o .cursor/rules/adhd-mode.mdc https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/rules/cursor/adhd-mode.mdcGitHub Copilot — .github/copilot-instructions.md
mkdir -p .github
curl -o .github/copilot-instructions.md https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/rules/copilot/copilot-instructions.mdWindsurf — .windsurf/rules/ (modern format)
mkdir -p .windsurf/rules
curl -o .windsurf/rules/adhd-mode.md https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/rules/windsurf/rules/adhd-mode.mdGemini CLI — GEMINI.md
curl -O https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/rules/gemini/GEMINI.mdTip: configure Gemini CLI to read AGENTS.md instead by adding {"context":{"fileName":"AGENTS.md"}} to .gemini/settings.json.
Codex — AGENTS.md (same as Option 1)
Codex reads AGENTS.md natively. No extra file needed.
Claude Code — CLAUDE.md or SKILL.md plugin
curl -O https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/CLAUDE.mdFor the on-demand plugin format (invoke-per-session), use claude/SKILL.md from this repo, or install the original i-have-adhd.
| Tool | Primary (AGENTS.md) | Per-tool file | Always-on? |
|---|---|---|---|
| Codex | Yes (native) | — | Yes |
| Cursor | Yes (native) | .cursor/rules/adhd-mode.mdc |
Yes (alwaysApply: true) |
| GitHub Copilot (cloud agent) | Yes (native) | — | Yes |
| GitHub Copilot (IDE chat) | — | .github/copilot-instructions.md |
Yes |
| Windsurf | Yes (experimental) | .windsurf/rules/adhd-mode.md |
Yes (trigger: always_on) |
| Gemini CLI | Yes (with config) | GEMINI.md |
Yes |
| Aider, Zed, Factory, Jules, Devin, Amp | Yes (native) | — | Yes |
| Claude Code | Via CLAUDE.md → @AGENTS.md |
SKILL.md (plugin) |
Yes (import) / On invocation (plugin) |
What is adhd-mode?
adhd-mode is a set of 10 output rules that make AI coding assistants produce ADHD-friendly output. Instead of burying the answer behind preamble, tangents, and pleasantries, the AI leads with the next action, numbers multi-step tasks, and ends with one concrete next step. It works by placing a single AGENTS.md file in your repository that 20+ AI coding tools read automatically.
How is this different from i-have-adhd?
i-have-adhd is a skill you invoke each session (/i-have-adhd in Claude Code, $i-have-adhd in Codex). adhd-mode is always on — you paste one file and every response is ADHD-shaped from the first message. It also covers tools the original doesn't (Copilot IDE chat, Windsurf modern format, Gemini CLI, Aider, Zed). The 10 rules are the same, credited and MIT-licensed. See docs/COMPARISON.md for the full breakdown.
Do I need an ADHD diagnosis to use this?
No. The rules help anyone who wants action-first, no-filler AI output. "ADHD-friendly" describes the output shape, not a medical requirement.
Which tools are supported?
Codex, Cursor, GitHub Copilot (cloud agent + IDE), Windsurf, Gemini CLI, Claude Code, Aider, Zed, Factory, Jules, Devin, Amp, and 20+ other tools that read AGENTS.md natively. See the table above.
How do I turn it off?
Say "stop adhd mode" or "normal mode" in any chat. The AI returns to its default output style for the rest of the session.
Will this slow down my AI or waste tokens?
The rules are ~2KB. Every tool loads them as part of the system prompt — negligible token cost. The output is shorter (no preamble, no closers), so responses actually use fewer tokens.
Can I customize the rules?
Yes. Fork the repo, edit scripts/build.py (the single source of truth), run python scripts/build.py to regenerate all files. See CONTRIBUTING.md.
Is this free?
Yes. MIT licensed. The rules are adapted from i-have-adhd (also MIT).
All rule files are generated from a single source (scripts/build.py). CI asserts parity — the 10 rules are identical across every file. To customize: fork, edit scripts/build.py, run it, commit, push.
To add a new tool: see CONTRIBUTING.md and the new-tool issue template.
The original plugin gives you the same 10 rules on demand. Use the original if you like invoking per session; use adhd-mode if you want it always on. Same rules, same author credited. They complement each other.
See ROADMAP.md for planned tools, features, and community milestones.
MIT. The 10 output rules are adapted from ayghri/i-have-adhd (MIT).