How to Disable or Override the *i-have-adhd* Skill: "Stop ADHD Mode" and Three Other Methods

The i-have-adhd skill can be disabled via four mechanisms: in-session voice commands ("stop adhd mode" or "normal mode"), CLI plugin disable commands, deleting the persistent flag file for always-on installs, or simply not invoking it due to disable-model-invocation: true metadata.

The i-have-adhd skill is designed to stay active throughout a session once triggered, applying ADHD-friendly formatting rules to AI responses. However, the ayghri/i-have-adhd repository provides multiple pathways to override or permanently disable this behavior depending on your runtime environment and installation type.

In-Session Command: "Stop ADHD Mode" or "Normal Mode"

The fastest way to disable the skill mid-conversation is with a direct voice command.

According to skills/i-have-adhd/SKILL.md, saying "stop adhd mode" or "normal mode" triggers an immediate override:

"Turn them off only when the reader says 'stop adhd mode' or 'normal mode'"【5†L19-L20】

The skill confirms deactivation in a single line and reverts to standard response formatting for the remainder of the conversation. This does not affect future sessions—only the current chat.

# Example interaction

User: stop adhd mode
Assistant: ✅ ADHD mode disabled (returning to normal style)

Plugin-Level Disable via CLI Commands

For runtimes using plugin management, you can disable the skill entirely for all future sessions. The exact command depends on your runtime harness:

• agy plugin disable i-have-adhd — for AGY-based runtimes
• claude plugin disable i-have-adhd — for Claude-based runtimes

These commands are documented in INSTALL.md at lines 31-33: "Or keep it installed and turn it off: agy plugin disable i-have-adhd"【6†L31-L33】. Unlike the in-session command, this persists across restarts until explicitly re-enabled.

Delete the Persistent Flag File (Always-On Installations)

When the skill is configured as always-on, the runtime creates a small flag file that triggers automatic injection on every session start. Removing this file permanently disables the always-on behavior.

As noted in INSTALL.md at lines 430-433: "delete the flag to turn always-on off for good"【6†L430-L433】.

The actual path varies by runtime, but typical locations include:

# Example paths (verify your runtime's configuration)

rm ~/.config/claude/plugins/i-have-adhd.enabled
rm ~/.config/agy/plugins/i-have-adhd.enabled

The flag-checking and injection logic is implemented in .opencode/plugins/i-have-adhd.mjs, which reads this file each turn to determine whether to apply ADHD formatting rules.

Disable Implicit Invocation via Metadata Flag

The skill's metadata in SKILL.md explicitly sets disable-model-invocation: true at lines 4-5【5†L4-L5】. This prevents automatic activation by the model—you must explicitly invoke the skill for it to take effect.

Invocation syntax varies by platform:

• Claude/Qwen: /i-have-adhd
• Codex: $i-have-adhd

By simply not using these invocation commands, you effectively disable the skill without any configuration changes. This is the most passive override mechanism available.

Code Reference: Hook Declaration

The always-on behavior is declared in hooks/hooks.json, which registers the runtime hook that checks for the flag file. When present, the hook in .opencode/plugins/i-have-adhd.mjs injects the ADHD formatting rules into each response.

Summary

Method	Scope	Use Case
"Stop adhd mode" / "normal mode"	Current session only	Quick temporary override mid-conversation
agy/claude plugin disable	All future sessions	Permanent disable without uninstalling
Delete flag file	Always-on installations	Permanent disable for auto-triggered setups
Don't invoke	Per-conversation	Passive avoidance when disable-model-invocation: true

Frequently Asked Questions

What happens if I say "stop adhd mode" in the middle of a response?

The skill immediately terminates its formatting rules for the current session. Any partial response completes without ADHD-specific structure, and subsequent responses use default styling. The change is instant and confirmed by a single-line status message.

Can I re-enable the skill after disabling it with "stop adhd mode"?

Yes. Since this command only affects the current session, you can restart the conversation or explicitly invoke the skill again using /i-have-adhd (Claude/Qwen) or $i-have-adhd (Codex) to reactivate ADHD mode.

Where exactly is the always-on flag file located?

The path is runtime-specific and not hardcoded in the repository. Check your runtime's plugin directory—common patterns include ~/.config/[runtime]/plugins/i-have-adhd.enabled or paths defined in your environment's plugin configuration. The INSTALL.md file provides runtime-specific guidance.

Does disable-model-invocation: true prevent the skill from working entirely?

No. This metadata flag only prevents automatic activation by the AI model. The skill functions normally when you explicitly invoke it with the platform-specific command. This design gives you full control over when ADHD formatting applies.

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