How to Invoke the i-have-adhd Skill Across Different AI Coding Assistants

The i-have-adhd skill is a portable output-shaping module that integrates with Claude Code, Codex, Gemini, and other AI assistants through marketplace installation and explicit commands like /i-have-adhd or implicit task detection.

The i-have-adhd repository (ayghri/i-have-adhd) provides a declarative skill interface that reformats AI-generated responses into ADHD-friendly structures. When you invoke this skill across different coding assistants, the underlying language model follows a strict ruleset defined in skills/i-have-adhd/SKILL.md to deliver action-first, numbered outputs without unnecessary fluff. Because the architecture relies on portable agent adapters rather than platform-specific code, you can activate consistent formatting behavior across any assistant that supports the skill marketplace protocol.

Understanding the i-have-adhd Skill Architecture

Before invoking the skill, it helps to understand how the repository structures its components to ensure portable activation across AI assistants.

Core Skill Definition

The behavioral contract resides in skills/i-have-adhd/SKILL.md. This file encodes ten explicit formatting rules—such as action-first headers and numbered sequential steps—that any host agent must obey when the skill is active. Because the rules are declarative, the skill requires no runtime logic changes to function across different large language models.

Agent Adapters

Each AI platform connects through tiny configuration files located in skills/i-have-adhd/agents/. For example, OpenAI-compatible agents (including Claude Code) use agents/openai.yaml, while Google Gemini uses agents/gemini.toml. These adapters specify the skill's display name, default prompt injection, and whether implicit invocation is permitted.

Marketplace Manifest

The plugin.json file at the repository root serves as the discovery mechanism. Assistants read this manifest to load the skill bundle without requiring a local clone, enabling seamless registration through marketplace CLI commands.

Installing and Invoking in Claude Code

To invoke the i-have-adhd skill in Claude Code, you must first register the repository in the assistant's marketplace and install the plugin bundle.

Run the following commands in your terminal:

claude plugin marketplace add ayghri/i-have-adhd
claude plugin install i-have-adhd@i-have-adhd

Once installed, trigger the skill explicitly by typing the command in any chat window:

/i-have-adhd

Claude will now prepend the rules from SKILL.md to its context, ensuring every subsequent response uses action-first formatting and numbered steps. According to the agents/openai.yaml configuration, the skill also supports implicit invocation—when allow_implicit_invocation is set to true, Claude automatically applies the skill to multi-step coding requests without requiring the /i-have-adhd prefix.

Installing and Invoking in Codex

The Codex CLI follows a similar marketplace pattern but uses slightly different command syntax. Register the repository and install the skill using:

codex plugin marketplace add ayghri/i-have-adhd --ref main
codex plugin add i-have-adhd@i-have-adhd

After installation, force the next response to obey ADHD-friendly formatting by typing:

$i-have-adhd

The $i-have-adhd command acts as an explicit trigger. Like Claude Code, Codex monitors your task stream for patterns that indicate multi-step work; when detected, it applies the skill automatically if implicit invocation is enabled in the adapter configuration.

Supporting Gemini and Other Assistants

For Google Gemini and other compatible agents, the repository includes agents/gemini.toml. This adapter follows the same metadata schema as the OpenAI variant, declaring the default prompt and invocation triggers. After adding ayghri/i-have-adhd to Gemini's marketplace using its specific CLI or GUI workflow, you can invoke the skill using the command defined in that adapter (typically /i-have-adhd).

Adding support for new assistants requires only creating a new adapter file that follows the host's plugin schema, because the core logic in SKILL.md never changes.

How the Skill Shapes Output at Runtime

When active, the assistant injects the default prompt defined in the relevant agent adapter into the LLM request pipeline. The model then formats its output according to the rules in SKILL.md, producing structured responses that prioritize clarity and actionability.

The skill remains active across turns until you send a stop phrase such as "stop adhd mode" or "normal mode", at which point the assistant returns to standard response formatting.

Summary

• The i-have-adhd skill from ayghri/i-have-adhd uses a declarative ruleset in skills/i-have-adhd/SKILL.md to enforce action-first, numbered output formatting.
• Agent adapters (agents/openai.yaml, agents/gemini.toml) bridge the skill to specific AI platforms without modifying core logic.
• Install via marketplace CLI using claude plugin marketplace add or codex plugin marketplace add --ref main, followed by the install command.
• Invoke explicitly using /i-have-adhd (Claude) or $i-have-adhd (Codex), or rely on implicit invocation for multi-step coding tasks.
• Deactivate the skill by saying "stop adhd mode" or "normal mode".

Frequently Asked Questions

What file contains the actual formatting rules for the i-have-adhd skill?

The formatting rules reside in skills/i-have-adhd/SKILL.md. This file contains the declarative behavioral contract—including the ten specific rules for action-first headers and numbered steps—that the AI assistant must follow when the skill is active.

Can I use the i-have-adhd skill without typing the invocation command every time?

Yes. The agents/openai.yaml configuration includes allow_implicit_invocation: true, which permits the assistant to automatically apply the skill when it detects multi-step coding requests. Once installed, the skill activates implicitly for suitable tasks without requiring /i-have-adhd or $i-have-adhd prefixes.

How do I stop the i-have-adhd skill once I've activated it?

To deactivate the skill and return to standard AI output formatting, send the stop phrase "stop adhd mode" or "normal mode" in your chat. The assistant will cease prepending the skill's ruleset to subsequent requests immediately.

Where should I place new adapter files if I want to add support for another AI assistant?

Create a new adapter file (YAML, TOML, or JSON depending on the target platform's requirements) inside the skills/i-have-adhd/agents/ directory. The adapter should declare the skill's display name, default prompt, and invocation settings according to the new assistant's plugin schema, while referencing the unchanged rules in SKILL.md.

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