Drop one file in your repo. Every AI coding tool — Cursor, Copilot, Windsurf, Gemini, Codex, Claude, and 20+ more — stops burying the answer and starts leading with action.clarity.results. Always on. No slash command.
Every AI coding assistant wastes your time with filler:
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 atsrc/auth.ts, theverifyTokenfunction seems to be using an older API... [47 lines] ...Hope this helps! Let me know if you want to dig deeper.
Runnpm install jsonwebtoken@latest, then editsrc/auth.ts:42.
1. Opensrc/auth.ts
2. ReplaceverifyToken(lines 42-58)
3. Runnpm test -- auth.spec.ts
Next: paste the first failing line.
Based on 324 HN comments about i-have-adhd (42k+ stars), v2.0 fixes the top complaints:
#1 complaint: rules fade after turns. adhd-mode ships re-injection instructions as an always-on output style.
#3 complaint: estimates are "inept." Replaced with Quick / Medium / Large.
#2 complaint: "8.7k lines, 59 files." adhd-mode is one file. 140 lines.
Also works as "action-first mode" for users who prefer neutral framing.
Install commands always pin exact GitHub raw URLs. No squatting risk.
AGENTS.md is a Linux Foundation standard. 20+ tools read it natively.
| # | Rule | What it does | ADHD challenge |
|---|---|---|---|
| 1 | Lead with the next action | First line = doable now | Task initiation |
| 2 | Number multi-step tasks | Numbered, bounded steps | Overwhelm |
| 3 | End with one next action | One thing under 2 min | Decision fatigue |
| 4 | Suppress tangents | Finish first, then second | Attention switching |
| 5 | Restate state every turn | "Step 3 of 5 done" | Working memory |
| 6 | Scope by shape | Quick / Medium / Large | Fabricated estimates |
| 7 | Make wins visible | Show what works | Dopamine deficit |
| 8 | Matter-of-fact errors | Cause + fix, no "uh oh" | Emotional dysregulation |
| 9 | Cap lists at 5 | Now vs later | Choice paralysis |
| 10 | No preamble, no closers | Start with answer | Scrolling fatigue |
curl -O https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/AGENTS.md
mkdir -p .cursor/rules
curl -o .cursor/rules/adhd-mode.mdc https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/rules/cursor/adhd-mode.mdcUses alwaysApply: true. Full guide →
mkdir -p .github
curl -o .github/copilot-instructions.md https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/rules/copilot/copilot-instructions.mdmkdir -p .windsurf/rules
curl -o .windsurf/rules/adhd-mode.md https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/rules/windsurf/rules/adhd-mode.mdcurl -O https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/rules/gemini/GEMINI.mdCodex reads AGENTS.md natively. Full guide →
curl -O https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/CLAUDE.md
curl -O https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/AGENTS.mdcurl -sSL https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/install/install.sh | bash
| Feature | i-have-adhd | adhd-mode |
|---|---|---|
| Activation | Invoke per session | Always on |
| Tools | Claude, Codex, Cursor | 20+ (AGENTS.md) |
| Rule decay fix | No | Anti-decay |
| Time estimates | Fabricated | Scope by shape |
| Repo size | 8.7k lines | 140 lines |
| Copilot IDE | No | Yes |
| Windsurf modern | No | Yes |
| License | MIT | MIT |
adhd-mode ships 10 output rules making AI coding assistants produce ADHD-friendly output. One AGENTS.md file works in 20+ tools including Cursor, Copilot, Windsurf, Gemini CLI, Codex, and Claude Code. Rules adapted from i-have-adhd (MIT).
i-have-adhd is invoked per session. adhd-mode is always on. It also fixes rule decay, fabricated time estimates, and bloat. Covers Copilot, Windsurf, and Gemini that the original doesn't.
Run: curl -O https://raw.githubusercontent.com/DevAnimecx/adhd-mode/main/AGENTS.md then restart your editor.
Codex, Cursor, Copilot, Windsurf, Gemini CLI, Claude Code, Aider, Zed, and 20+ others via AGENTS.md.
No. Anti-decay mechanism with re-injection. Say 'adhd mode' again to re-apply.
No. Works as 'action-first mode' for anyone.
Replaced with scope-by-shape: Quick, Medium, Large.
Yes. MIT licensed. No paid features.
One file. 30 seconds. Every AI tool.