Runtime Entry Points for Gemini and Qwen Integrations in i-have-adhd

The i-have-adhd project uses separate JSON manifest files—gemini-extension.json and qwen-extension.json—as runtime entry points to register ADHD-friendly response rules in Gemini CLI and Qwen Code.

The i-have-adhd repository provides LLM-agnostic ADHD coaching rules that integrate with multiple AI coding assistants. Understanding the runtime entry points for Gemini and Qwen integrations reveals how these platforms discover and load the skill without modifying core runtime code.

How Runtime Entry Points Work

Both integrations rely on declarative manifest files that the respective CLIs scan at startup. These manifests declare the skill's metadata, entry commands, and context sources, allowing the runtimes to expose the /i-have-adhd command natively.

Gemini CLI Integration

For Gemini CLI, the runtime entry point is [gemini-extension.json](https://github.com/ayghri/i-have-adhd/blob/main/gemini-extension.json). This manifest registers the skill as a native Gemini extension.

Key fields in the manifest:

• name and version – Identify the extension in the Gemini ecosystem
• contextFileName – Points to GEMINI.md, which contains the skill description and rules consumed by the LLM
• description – Human-readable summary shown in extension listings

When Gemini CLI initializes, it loads this manifest, reads the referenced [GEMINI.md](https://github.com/ayghri/i-have-adhd/blob/main/GEMINI.md), and makes the /i-have-adhd command available as a plug-in.

Qwen Code Integration

For Qwen Code, the runtime entry point is [qwen-extension.json](https://github.com/ayghri/i-have-adhd/blob/main/qwen-extension.json). This file exposes the skill as a native Qwen extension.

Key fields include:

• skills – Specifies the directory containing skill implementations (typically skills/i-have-adhd/)
• name, version, and description – Standard metadata for the Qwen marketplace

Qwen Code scans the skills directory at startup to locate [skills/i-have-adhd/SKILL.md](https://github.com/ayghri/i-have-adhd/blob/main/skills/i-have-adhd/SKILL.md), which holds the 10 ADHD-friendly response rules applied by the runtime.

Configuration and Skill Implementation

Both entry points ultimately resolve to the same core rules, but they reference different context files due to platform-specific formatting requirements.

Core Skill Definition

The actual behavioral logic resides in SKILL.md, which contains:

• 10 structured rules for ADHD-friendly responses (bullet points, clear headings, concise instructions)
• disable-model-invocation: true flag, ensuring the skill remains inactive until explicitly invoked
• Deactivation triggers: "stop adhd mode" or "normal mode"

File Path Reference

Platform	Entry Point Manifest	Context File	Implementation
Gemini	gemini-extension.json	GEMINI.md	Loaded as context at runtime
Qwen	qwen-extension.json	skills/i-have-adhd/SKILL.md	Scanned from skills directory

Activating and Deactivating Skills

Once the runtime loads the manifest, users interact with the skill through standardized slash commands.

Gemini CLI Usage

$ gemini
> /i-have-adhd

Qwen Code Usage

$ qwen
> /i-have-adhd

Deactivation Commands

To return to standard behavior in either runtime:

> stop adhd mode

Or alternatively:

> normal mode

Summary

• Gemini integration uses gemini-extension.json as its runtime entry point, referencing GEMINI.md via the contextFileName field.
• Qwen integration uses qwen-extension.json, discovering rules in the skills/i-have-adhd/ directory.
• Both platforms expose the skill through the /i-have-adhd slash command.
• The skill respects the disable-model-invocation: true flag, remaining dormant until explicitly activated.
• Deactivation occurs by typing "stop adhd mode" or "normal mode" in either CLI.

Frequently Asked Questions

Where are the runtime entry points defined in the i-have-adhd repository?

The runtime entry points are defined in the repository root as gemini-extension.json for Gemini CLI and qwen-extension.json for Qwen Code. These JSON files contain the metadata required for each platform to discover and register the skill at startup.

Do Gemini and Qwen use the same skill implementation file?

No. While both implement the same 10 ADHD-friendly rules, Gemini consumes GEMINI.md (referenced via contextFileName), whereas Qwen reads skills/i-have-adhd/SKILL.md (discovered through the skills directory array). Both files contain identical behavioral logic but may differ in formatting optimized for each platform's context window handling.

Can the skill activate automatically when I start the CLI?

No. According to the disable-model-invocation: true flag defined in SKILL.md, the skill remains inactive by default. You must explicitly type /i-have-adhd in either Gemini CLI or Qwen Code to enable ADHD mode for that session.

What command deactivates the ADHD skill in both runtimes?

Typing stop adhd mode or normal mode disables the skill and returns the LLM to standard response patterns. This command works identically in both Gemini CLI and Qwen Code environments.

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