i-have-adhd: 10 Rules to Make AI Coding Agents Concise and Action-First

What It Does

The i-have-adhd skill tackles a common frustration: AI coding agents often respond with verbose explanations, polite filler, and background context before delivering the actual code change or command. The project provides a single rule file ( skills/i-have-adhd/SKILL.md) that reshapes agent output into a terse, executable format.

Before vs. After Comparison

The README includes a real-world example. When asked about updating a JWT verification function:

Before (default agent): "Great question! Let me think about this. Your auth flow has a few moving pieces: the middleware, the token verification, and the cookie handling. Looking at src/auth.ts , the verifyToken function (around lines 42-58) seems to be using an older jsonwebtoken API. 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." After (with i-have-adhd): "Run npm install jsonwebtoken@latest , then edit src/auth.ts:42 ." Open src/auth.ts Replace verifyToken (lines 42–58) with the snippet below Run npm test -- auth.spec.ts Next: paste the first failing line if any test fails."

The difference is immediate: the second response puts the exact command and file location in the first sentence, followed by a numbered step list and a clear next action.

Key Rules (Selected)

Lead with the next action. The first sentence must be a concrete step the user can execute.

Restate state every turn. Each reply includes progress context (e.g., "Step 3 of 5 completed") so interrupted sessions resume without scrolling.

Specific time estimates. Use "3 minutes" instead of "a moment."

No preamble. No recap. No closers. Eliminates greetings, summaries, and "hope this helps" sign-offs.

Additional rules: number all steps, limit lists to five items, report errors factually without apology.

Installation & Activation

The plugin must be explicitly invoked after installation; it does not activate by default.

Claude Code / Qwen Code: type /i-have-adhd in chat.

Codex: type $i-have-adhd.

Persistent activation (Claude Code): run touch ~/.claude/.i-have-adhd-always; a SessionStart hook loads the rules from the first message. Disable anytime with "stop adhd mode".

One-line install commands:

claude plugin marketplace add ayghri/i-have-adhd
claude plugin install i-have-adhd@i-have-adhd

Other environments:

Cursor: npx skills add ayghri/i-have-adhd Gemini CLI:

gemini extensions install https://github.com/ayghri/i-have-adhd

Kimi Code CLI: use /plugins → Custom → paste repo URL

Zed, OpenCode, Copilot, Pi: instructions in the repo's INSTALL.md (includes Chinese version).

Exceptions (Safety Over Brevity)

The skill defines six override conditions:

If the user explicitly asks for an explanation ("explain the principle"), the agent provides detail but still omits preamble/closer.

Destructive commands ( rm -rf, force push) require confirmation before execution.

After three consecutive failed fixes, the agent stops and states which assumption may be wrong instead of guessing further.

Author's Take

i-have-adhd doesn't make the model smarter; it makes the output executable. The distinction matters: intelligence and executability are often conflated but are separate. The tool suits developers who use Claude Code, Cursor, or Codex daily and are tired of wading through polite filler. It is not suited for learners who need concepts broken down step by step — the rules will feel too dry. The author notes that any agent reply exceeding three lines without a concrete action now feels intolerable.

Source

Open-source repository:

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