How to Make the i-have-adhd Skill Persistent Across All Sessions Without Reinvoking It

To make the i-have-adhd skill persistent across all sessions, add the skill's core rule set to your AI agent's persistent configuration file so it loads automatically at startup instead of requiring manual activation.

The i-have-adhd skill by ayghri/i-have-adhd is designed for on-demand activation via the /i-have-adhd slash command. However, you can eliminate the need to reinvoke it every session by integrating its rules into your agent's persistent configuration. This guide covers the exact steps for every supported harness, with file paths and commands straight from the source code.

Why Persistence Requires Configuration Changes

The skill's core rules live in skills/i-have-adhd/SKILL.md. When you type /i-have-adhd, the harness loads this file into context. To make this automatic, you must place an equivalent rule block where your harness reads before the first user message. Each harness exposes a different persistent location for this purpose.

Claude Code: Enable Always-On via Flag File

Claude Code uses a hook-based system. The hooks/hooks.json file defines a SessionStart hook that checks for ~/.claude/.i-have-adhd-always and runs always-on.sh when present.

# Create the empty flag file — no content needed

touch ~/.claude/.i-have-adhd-always

# Restart Claude Code to activate

Once created, the hook automatically loads the full ruleset from skills/i-have-adhd/SKILL.md before your first message.

Cursor: Add to User Rules or Project Rules

Cursor reads persistent rules from two locations. Choose based on your needs:

Global (all projects): Settings → Rules → User Rules

Project-specific: .cursor/rules/ directory with alwaysApply: true

Paste this block:

## Output style

The reader has ADHD. Shape every response so it can be acted on:

1. Lead with the answer or next action: command, path, or snippet first.
2. Number multi‑step work; one bounded action per step.
3. End with one next action doable in under two minutes.
4. Finish the current issue before raising a new one.
5. Restate progress each turn ("step 3 of 5 done").
6. Give time estimates in concrete units, never "a bit".
7. After a change, show what now works.
8. Errors: state location, cause, and fix. No drama.
9. Cap lists at 5 items.
10. No preamble, no recaps, no closers.

This matches the INSTALL.md instructions for Cursor persistence.

OpenCode: Append to AGENTS.md

OpenCode loads ~/.config/opencode/AGENTS.md at every session start.

cat <<'EOF' >> ~/.config/opencode/AGENTS.md

## Output style

The reader has ADHD. Shape every response so it can be acted on:

1. Lead with the answer or next action: command, path, or snippet first.
2. Number multi‑step work; one bounded action per step.
3. End with one next action doable in under two minutes.
4. Finish the current issue before raising a new one.
5. Restate progress each turn ("step 3 of 5 done").
6. Give time estimates in concrete units, never "a bit".
7. After a change, show what now works.
8. Errors: state location, cause, and fix. No drama.
9. Cap lists at 5 items.
10. No preamble, no recaps, no closers.
EOF

Restart OpenCode to load the new rules.

Gemini CLI: Install as Extension

The Gemini CLI extension ships with GEMINI.md containing the always-on block. Installing the extension handles persistence automatically.

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

The extension copies GEMINI.md to the appropriate location, making the skill active from message 1.

GitHub Copilot / Codex: Project-Level Instructions

For Codex and Copilot-enabled editors, add the rule block to:

.github/copilot-instructions.md

Commit this file to your repository. Codex reads it at the start of every chat session in that project.

Hermes: AGENTS.md or SOUL.md

Hermes supports two persistence modes:

• Project-scoped: AGENTS.md in your working directory
• Global sessions: SOUL.md

Add the same markdown block to either file based on your desired scope.

Pi: Project AGENTS.md

Pi reads AGENTS.md from the project root. Place the rule block there to activate the skill for all sessions in that project.

Zed: Personal Configuration

Zed loads agent rules from:

~/.config/zed/AGENTS.md

Append the skill's rule block to this file for persistent activation across all Zed sessions.

Generic Harnesses: Identify Your Persistent Rules File

For any Agent Skills-compatible harness (Cursor, OpenCode, AMP, or similar):

1. Locate the harness's persistent rules location (consult its documentation)
2. Copy the 10-rule markdown block from skills/i-have-adhd/SKILL.md
3. Paste into that file and restart

The INSTALL.md file in the repository contains harness-specific guidance if your tool isn't listed above.

How the Always-On Mechanism Works

When properly configured, the harness reads the rule block at session initialization — before any user message. The rules from skills/i-have-adhd/SKILL.md are injected into context immediately, transforming every response to follow ADHD-optimized formatting without requiring the /i-have-adhd trigger.

This differs from on-demand activation where:

• The user must remember to invoke the skill
• The rules load mid-conversation
• Prior context lacks the formatting constraints

Persistent configuration ensures zero-friction usage: every session starts with the constraints already active.

Summary

• Claude Code: Create ~/.claude/.i-have-adhd-always and restart
• Cursor: Paste rules into User Rules or .cursor/rules/ with alwaysApply: true
• OpenCode/Zed: Append to ~/.config/{harness}/AGENTS.md
• Gemini CLI: Install the extension once
• Codex: Add to .github/copilot-instructions.md
• Hermes/Pi: Use project AGENTS.md or global SOUL.md
• All methods source rules from skills/i-have-adhd/SKILL.md as documented in INSTALL.md

Frequently Asked Questions

What happens if I use both the flag file and paste rules manually?

The skill activates through whichever mechanism your harness supports. In Claude Code, the flag file triggers the full ruleset from SKILL.md, so manual pasting is redundant. For other harnesses without hook systems, manual placement in the persistent file is the only supported path.

Does persistent activation slow down session startup?

No measurable impact. The rule block is plain text (~500 bytes) loaded once at initialization. The hooks/hooks.json implementation in Claude Code and similar mechanisms in other harnesses run synchronously before message processing begins.

Can I disable persistence without uninstalling the skill?

Yes. For Claude Code, delete ~/.claude/.i-have-adhd-always. For file-based harnesses, remove or comment the block from your AGENTS.md or rules file. The skill remains available for on-demand use via /i-have-adhd in all cases.

Where do the 10 formatting rules come from?

The 10 rules are defined in skills/i-have-adhd/SKILL.md and replicated across all harness-specific installation paths. Each rule targets a specific ADHD communication need: bounded actions, concrete time estimates, progress restatement, and elimination of cognitive overhead like preambles and recaps.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}