What Does the disable-model-invocation Flag Do in i-have-adhd?

The disable-model-invocation flag prevents the i-have-adhd skill from automatically activating when the repository loads, requiring users to explicitly invoke it via commands like /i-have-adhd before its specialized behavior takes effect.

The disable-model-invocation flag is a metadata directive in the ayghri/i-have-adhd repository that controls when the skill's "action-first" output style applies. Defined in the skill's manifest file, this flag ensures that AI assistants such as GitHub Copilot and Claude Code do not silently modify their responses until the user explicitly requests the skill's assistance.

Where the Flag Is Defined

In skills/i-have-adhd/SKILL.md, the flag appears at line 4 within the skill's YAML frontmatter:

---
name: i-have-adhd
description: ADHD-friendly output formatting
disable-model-invocation: true
license: MIT
---

Setting disable-model-invocation: true signals to the hosting AI assistant that this skill should remain dormant during standard interactions, waiting for explicit user activation rather than applying automatically.

How Explicit Invocation Works

When disable-model-invocation is enabled, the skill's behavior is applied only after explicit user action. Instead of automatically formatting every response, the model maintains its default style until you type the activation command.

To invoke the skill manually, use:

/i-have-adhd

This command activates the skill temporarily, applying its action-first output style to subsequent interactions. According to the source code documentation in INSTALL.md, both Copilot and Claude Code respect this flag identically: "nothing applies until you invoke the skill."

OpenAI Configuration Equivalent

For environments using OpenAI-specific configurations, the repository provides an equivalent control in skills/i-have-adhd/agents/openai.yaml at line 7:

policy:
  allow_implicit_invocation: false

This policy.allow_implicit_invocation: false setting mirrors the behavior of disable-model-invocation: true, ensuring the skill is never implicitly activated regardless of the hosting platform's specific implementation.

Why Use disable-model-invocation?

Controlling when the skill activates provides several advantages:

• User agency: Prevents the AI from applying ADHD-specific formatting when you're asking general coding questions
• Context switching: Allows normal conversational assistance until you specifically need the structured, action-first output style
• Cross-platform consistency: Ensures predictable behavior across Copilot, Claude Code, and other AI assistants that parse the skill metadata

Summary

• The disable-model-invocation flag in skills/i-have-adhd/SKILL.md stops automatic skill activation at line 4
• Users must explicitly invoke the skill with /i-have-adhd to enable its specialized behavior
• OpenAI configurations use policy.allow_implicit_invocation: false in agents/openai.yaml for equivalent control
• The flag is respected by GitHub Copilot, Claude Code, and other compatible assistants according to INSTALL.md

Frequently Asked Questions

How do I activate the i-have-adhd skill if disable-model-invocation is set to true?

Type /i-have-adhd in your chat interface. This explicit command overrides the disabled state and applies the skill's action-first formatting to the current conversation, as documented in the repository's installation guide.

What happens if I remove the disable-model-invocation flag from SKILL.md?

Removing the flag or setting it to false allows the AI assistant to implicitly invoke the skill whenever it detects relevant context, potentially applying the action-first style to all responses without requiring the explicit /i-have-adhd command.

Does the disable-model-invocation flag work with all AI assistants?

According to the INSTALL.md documentation in the ayghri/i-have-adhd repository, Copilot and Claude Code fully respect this flag. Other assistants that follow the skill metadata specification should also honor it, though behavior may vary by platform implementation.

Where can I find the OpenAI-specific configuration for this setting?

The equivalent setting for OpenAI agents is located in skills/i-have-adhd/agents/openai.yaml at line 7, using policy.allow_implicit_invocation: false to prevent automatic skill activation in OpenAI-compatible environments.

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