How to Invoke the i-have-adhd Skill on Demand: Complete Runtime Activation Guide

To invoke the i-have-adhd skill on demand, type the runtime-specific activation command—/i-have-adhd in Claude Code, $i-have-adhd in Codex, or /i-have-adhd in most other supported environments—which immediately applies ADHD-friendly formatting rules to your current session until you say "stop adhd mode."

The i-have-adhd skill in the ayghri/i-have-adhd repository is an Agent Skill that modifies AI output structure for improved readability. Unlike standard skills that load automatically, this skill requires explicit invocation due to its disable-model-invocation: true metadata flag defined in skills/i-have-adhd/SKILL.md, ensuring users only activate accessibility modifications when specifically needed.

Why Explicit Invocation Is Required

The skill’s activation architecture relies on a deliberate opt-in mechanism. In skills/i-have-adhd/SKILL.md (lines 1-5), the front-matter includes disable-model-invocation: true, which instructs every runtime harness to ignore the skill during automatic loading phases. This design prevents ADHD formatting rules from applying to unrelated conversations unless you explicitly trigger them.

According to the source code, this flag forces the runtime to wait for a slash command (or dollar-sign command in Codex) that matches the skill’s name exactly. Once triggered, the harness loads the rule set from SKILL.md and applies text-processing modifications for the duration of the session.

Installation Prerequisites

Before invoking the skill, you must install it for your specific runtime as documented in INSTALL.md. Each environment provides specialized installation commands:

# Claude Code

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

# Codex

codex plugin marketplace add ayghri/i-have-adhd --ref main
codex plugin add i-have-adhd@i-have-adhd

# Gemini CLI

curl -fsSL https://raw.githubusercontent.com/ayghri/i-have-adhd/main/skills/i-have-adhd/agents/gemini.toml \
  -o ~/.gemini/commands/i-have-adhd.toml

After installation, the skill remains dormant—available but inactive—until you issue the specific invocation command for your environment.

Runtime-Specific Invocation Commands

The exact syntax to invoke the i-have-adhd skill varies by runtime environment. Reference INSTALL.md for the complete mapping of entry points.

Claude Code

In Claude Code, use the slash command syntax:

/i-have-adhd

The command routes through Claude’s plugin system, loading the rules defined in skills/i-have-adhd/SKILL.md and immediately reformating subsequent outputs for ADHD-friendly consumption.

Codex

Codex requires a dollar-sign prefix rather than a slash:

$i-have-adhd

This distinct syntax prevents command collision with Codex’s built-in slash command namespace while activating the same i-have-adhd rule set.

Gemini CLI

For Gemini CLI, use the standard slash route after installing the command definition:

/i-have-adhd

Alternatively, installing the extension via the gemini.toml configuration file in ~/.gemini/commands/ enables the command route persistence across sessions.

OpenCode

OpenCode loads the skill as a plugin module defined in opencode.json. After cloning the repository to ~/.config/opencode/vendor/i-have-adhd and configuring the plugin path, invoke with:

/i-have-adhd

The activation triggers the JavaScript handler in .opencode/plugins/i-have-adhd.mjs, which processes the SKILL.md content through OpenCode’s extension API.

Pi

Pi supports both toggling and explicit state management:

/i-have-adhd        # Toggles the mode on

/i-have-adhd off    # Explicitly disables

This implementation reads from INSTALL.md’s Pi-specific section, providing granular control over the skill’s text-processing pipeline.

Other Runtimes

The following environments use standard slash-command invocation as documented in AGENTS.md:

• Oh My Pi (OMP): /i-have-adhd
• Qwen Code: /i-have-adhd
• Zed: /i-have-adhd (after adding the skill URL to Zed’s configuration)

Deactivating the Skill and Session Persistence

Once invoked, the skill remains persistently active for the entire session. The formatting rules apply to all subsequent AI responses until explicitly disabled.

To deactivate, use the termination phrases recognized by the runtime harness:

stop adhd mode

or

normal mode

These commands signal the runtime to unload the text-processing rules and return to standard output formatting. The session persistence mechanism is implemented in the runtime-specific handlers referenced in INSTALL.md under the How activation works section.

Configuring Always-On Activation

If you prefer the skill active by default without typing the invocation command each session, create a flag file in your runtime’s configuration directory. The hooks/hooks.json file defines the loading mechanism that checks for these flags.

For Claude Code, create:

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

When this flag file exists, the hook loader automatically applies the i-have-adhd rules at session startup, bypassing the need for manual /i-have-adhd invocation. Similar flag configurations exist for other runtimes as documented in INSTALL.md’s Always-on snippets.

Summary

• The i-have-adhd skill requires explicit invocation due to disable-model-invocation: true in skills/i-have-adhd/SKILL.md.
• Claude Code uses /i-have-adhd, while Codex requires $i-have-adhd.
• Most other runtimes (Gemini, OpenCode, Pi, OMP, Qwen Code, Zed) use the /i-have-adhd slash command.
• The skill remains active for the entire session until you say "stop adhd mode" or "normal mode".
• Create runtime-specific flag files (e.g., ~/.claude/.i-have-adhd-always) to enable automatic loading without manual invocation.

Frequently Asked Questions

What makes the i-have-adhd skill different from other agent skills?

Unlike standard agent skills that load automatically when a model detects relevant context, the i-have-adhd skill includes disable-model-invocation: true in its SKILL.md metadata. This flag forces runtimes to ignore the skill during automatic invocation phases, requiring users to explicitly activate it via command. This design ensures ADHD-specific formatting rules—such as structured bullet points and concise paragraphs—only apply when explicitly requested.

Can I use the skill across multiple runtimes simultaneously?

Yes, the skill is runtime-agnostic in its core definition but requires separate installation and invocation per environment. Since each runtime (Claude Code, Codex, Gemini CLI, etc.) maintains its own plugin namespace and configuration directory as mapped in AGENTS.md, you must install and invoke /i-have-adhd (or $i-have-adhd for Codex) in each active environment independently.

Why does the skill use disable-model-invocation: true?

The disable-model-invocation: true setting prevents the AI from automatically applying ADHD-friendly formatting to unrelated conversations where structured output might be unnecessary or disruptive. By requiring explicit invocation through commands like /i-have-adhd, the skill ensures accessibility modifications are intentional and context-appropriate, preserving standard formatting for general use while providing specialized support on demand.

How do I verify the skill is currently active?

After invoking the command (e.g., /i-have-adhd), the runtime loads the rules from skills/i-have-adhd/SKILL.md into the active session context. You can verify activation by observing that subsequent AI responses utilize ADHD-friendly formatting patterns—specifically, content structured with frequent line breaks, clear bullet hierarchies, and concise chunked information. To confirm the skill is processing, check that the runtime’s plugin status (documented in INSTALL.md per environment) lists i-have-adhd as an active extension.

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