# What Is `disable-model-invocation` in the i-have-adhd Skill? > Understand disable-model-invocation in the i-have-adhd skill. Learn how this metadata flag controls automatic loading, ensuring explicit user control. - Repository: [Ayoub Ghriss/i-have-adhd](https://github.com/ayghri/i-have-adhd) - Tags: deep-dive - Published: 2026-08-04 --- **The `disable-model-invocation` flag is a metadata property that prevents the i-have-adhd skill from automatically loading, requiring explicit user invocation via its command (e.g., `/i-have-adhd`).** The i-have-adhd skill is an open-source productivity tool designed to reshape AI output for readers with ADHD. According to the ayghri/i-have-adhd repository, the skill uses a specific control flag to govern when its formatting rules are applied. Understanding this mechanism is essential for users integrating the skill with Claude Code, Copilot, or similar AI harnesses. ## How `disable-model-invocation` Works in SKILL.md The flag is declared as a top-level metadata property in the skill's descriptor file. In [`skills/i-have-adhd/SKILL.md`](https://github.com/ayghri/i-have-adhd/blob/main/skills/i-have-adhd/SKILL.md), line 4 sets this value: ```yaml disable-model-invocation: true ``` This single line controls the skill's activation behavior. When present and set to `true`, the surrounding harness interprets this as an instruction to **ignore the skill's rules during normal operation**. The complete skill definition follows this pattern: ```yaml --- name: i-have-adhd description: 'Shape output for a reader with ADHD …' disable-model-invocation: true # ← prevents auto‑loading license: MIT --- ``` Without this flag (or with it set to `false`), the harness would apply the skill's output-shaping rules to every response automatically—behavior the skill author explicitly avoids. ## Why Explicit Invocation Is Required The i-have-adhd skill modifies how AI responses are structured: breaking content into scannable sections, using bold emphasis strategically, and limiting paragraph length. These transformations are **not universally desired**—they're only helpful when the user specifically needs ADHD-optimized formatting. The flag enforces this opt-in design: - **Default state:** The model receives no special instructions; responses follow standard formatting - **Activated state:** The user runs `/i-have-adhd` and the skill's rules immediately reshape subsequent output This mirrors Claude Code's skill invocation pattern, where specialized behaviors remain dormant until explicitly called. ## Platform-Specific Behavior The repository documents how major AI platforms respect this flag. ### Copilot Integration Per [`INSTALL.md`](https://github.com/ayghri/i-have-adhd/blob/main/INSTALL.md) around line 406, Copilot checks for `disable-model-invocation` before loading any skill. The platform's harness skips the i-have-adhd skill entirely during automatic skill detection, only surfacing it when the user types the invocation command. ### Claude Code vs. Codex Differences A later section in [`INSTALL.md`](https://github.com/ayghri/i-have-adhd/blob/main/INSTALL.md) (line 560) clarifies behavioral differences between platforms. Claude Code strictly honors the flag, while some codex implementations may vary in their default skill-loading behavior. The explicit `true` setting ensures consistent cross-platform dormancy. ## File Locations and Reference | File | Path | Purpose | |------|------|---------| | Primary skill definition | [`skills/i-have-adhd/SKILL.md`](https://github.com/ayghri/i-have-adhd/blob/main/skills/i-have-adhd/SKILL.md) | Contains the `disable-model-invocation` flag and core rules | | Cached copy | [`.cursor/skills/i-have-adhd/SKILL.md`](https://github.com/ayghri/i-have-adhd/blob/main/.cursor/skills/i-have-adhd/SKILL.md) | Mirror of the skill descriptor for Cursor IDE users | | Installation guide | [`INSTALL.md`](https://github.com/ayghri/i-have-adhd/blob/main/INSTALL.md) | Platform-specific integration instructions | All three locations maintain identical flag values to ensure consistent behavior across deployment contexts. ## Practical Activation Example To use the skill after installation: ```bash /i-have-adhd # user explicitly invokes the skill ``` This command overrides the `disable-model-invocation` restriction for that session, loading the ADHD-optimized formatting rules on demand. ## Summary - **`disable-model-invocation`** in [`skills/i-have-adhd/SKILL.md`](https://github.com/ayghri/i-have-adhd/blob/main/skills/i-have-adhd/SKILL.md) controls automatic skill loading - Set to `true` on line 4, it keeps the skill dormant until explicit invocation - Platforms like Copilot and Claude Code respect this flag per [`INSTALL.md`](https://github.com/ayghri/i-have-adhd/blob/main/INSTALL.md) documentation - Users activate the skill via `/i-have-adhd` command when needed - The cached copy at [`.cursor/skills/i-have-adhd/SKILL.md`](https://github.com/ayghri/i-have-adhd/blob/main/.cursor/skills/i-have-adhd/SKILL.md) maintains identical configuration ## Frequently Asked Questions ### What happens if `disable-model-invocation` is set to `false`? The harness would automatically apply the i-have-adhd skill's formatting rules to every AI response without requiring user invocation. This would force ADHD-optimized output even when standard formatting is preferred, defeating the skill's purpose as an on-demand productivity tool. ### Do I need to modify the flag to use the skill? No modification is necessary. The default `true` value is intentional—simply type `/i-have-adhd` to activate the skill when you need its formatting benefits. This preserves normal AI behavior as the default. ### Is `disable-model-invocation` a standard skill property? While not universal, this flag follows emerging conventions in AI assistant configuration. The ayghri/i-have-adhd repository aligns with Claude Code's skill ecosystem, where explicit invocation patterns are preferred for specialized output transformations. ### Where else is this flag documented beyond SKILL.md? The [`INSTALL.md`](https://github.com/ayghri/i-have-adhd/blob/main/INSTALL.md) file references the flag at lines 406 and 560, explaining how Copilot respects it and contrasting Claude Code's behavior with Codex implementations. These sections provide platform-specific context for developers integrating the skill.