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

The disable-model-invocation flag is a YAML metadata property defined in the skill descriptor that prevents automatic rule application until the user explicitly invokes the skill.

This flag controls activation behavior in the ayghri/i-have-adhd repository, ensuring the productivity-focused rules remain dormant during standard AI interactions. Located in the skill's frontmatter, the disable-model-invocation setting determines whether Claude Code, Copilot, or similar harnesses automatically load the output-shaping instructions.

Where the disable-model-invocation Flag Is Defined

The metadata property appears at line 4 of skills/i-have-adhd/SKILL.md:

---
name: i-have-adhd
description: 'Shape output for a reader with ADHD …'
disable-model-invocation: true   # prevents auto-loading

license: MIT
---

A cached copy of the skill definition exists at .cursor/skills/i-have-adhd/SKILL.md, maintaining identical configuration to the source file.

How disable-model-invocation Controls Skill Loading

When set to true, this property instructs the surrounding harness not to automatically apply the skill's rules. The model never receives the output-shaping instructions during standard chat sessions, keeping the skill dormant by default.

Explicit Invocation Requirements

To activate the skill, you must type the explicit command:

/i-have-adhd

This manual trigger temporarily loads the rules for that specific interaction. If the flag were set to false or omitted, the harness would automatically prepend the skill's instructions to every prompt, creating an "always-on" behavior.

How Platforms Respect the disable-model-invocation Flag

According to the installation documentation in INSTALL.md, both Copilot and Claude Code honor this metadata setting.

Copilot recognizes the flag and remains inactive until explicitly invoked (see INSTALL.md around line 406). Similarly, Claude Code distinguishes manually triggered skills from auto-loaded ones, as documented near line 560 of INSTALL.md, ensuring consistent behavior across platforms.

Summary

• The disable-model-invocation flag appears at line 4 of skills/i-have-adhd/SKILL.md and its cached copy at .cursor/skills/i-have-adhd/SKILL.md
• Setting it to true prevents automatic skill loading across Claude Code, Copilot, and Codex
• Users must explicitly invoke the skill using /i-have-adhd to activate its output-shaping rules
• Omitting the flag or setting it to false forces the skill rules to apply to every interaction

Frequently Asked Questions

What happens if I set disable-model-invocation to false?

If set to false or omitted entirely, the harness automatically applies the skill's output-shaping rules to every response. This creates an "always-on" behavior that contradicts the intended productivity focus of the i-have-adhd skill.

How do I know if the skill is active when disable-model-invocation is true?

The skill is only active immediately after you type /i-have-adhd in the chat interface. Once invoked, the AI applies the formatting rules for that specific interaction, returning to normal behavior afterward.

Does GitHub Copilot respect the disable-model-invocation flag?

Yes. According to INSTALL.md at line 406, Copilot recognizes the flag and keeps the skill dormant until explicitly called via the invocation command.

Where is the skill definition cached locally?

The repository includes a cached copy at .cursor/skills/i-have-adhd/SKILL.md that mirrors the canonical definition in skills/i-have-adhd/SKILL.md. Both files contain identical disable-model-invocation settings.

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