How the i-have-adhd Skill Modifies the System Prompt for ADHD-Friendly Output

The i-have-adhd skill transforms generic LLM responses into ADHD-friendly output by injecting a specific set of formatting rules into the system prompt via the $i-have-adhd placeholder expansion.

The ayghri/i-have-adhd repository implements a prompt-modifying skill that rewrites the system prompt to optimize every model response for readers with ADHD. By leveraging a placeholder variable and a dedicated rule file, the i-have-adhd skill ensures that outputs follow a strict action-first format without requiring users to manually prompt for structure each time.

The Mechanism: Prompt Injection via Placeholder Expansion

The skill operates through a three-stage pipeline that dynamically modifies the system prompt at runtime.

First, the complete behavioral rule set is defined in skills/i-have-adhd/SKILL.md. This file contains concrete constraints such as leading with the next action, numbering multi-step tasks, and ending with one concrete next action.

Second, the OpenAI agent configuration in skills/i-have-adhd/agents/openai.yaml declares a default prompt containing the placeholder $i-have-adhd. This placeholder acts as a marker where the rules will be inserted.

Third, when the user activates the skill via the /i-have-adhd command, the system harness replaces the $i-have-adhd placeholder with the full text content of SKILL.md. The resulting system prompt now contains the original directives plus the ADHD-specific constraints, causing all subsequent model generations to adhere to the formatting rules.

Core Configuration Files

SKILL.md – The Rule Engine

Located at skills/i-have-adhd/SKILL.md, this file houses the complete set of ADHD-specific output rules. The constraints include suppressing tangents, providing concrete time estimates, and structuring responses with numbered steps. When expanded, this content is injected directly into the active system prompt.

agents/openai.yaml – The Placeholder Definition

The agent configuration file defines how the skill interfaces with OpenAI models. It specifies the interface display name and, critically, includes the default_prompt field containing the $i-have-adhd variable. This YAML configuration ensures the placeholder is present in the prompt template before expansion occurs.

agents/gemini.toml – Cross-Platform Support

For Gemini-based agents, the repository provides skills/i-have-adhd/agents/gemini.toml, which mirrors the placeholder logic for Google's Gemini models. This ensures consistent ADHD-friendly formatting across different LLM providers.

Activation and Session Persistence

Users activate the skill by issuing the /i-have-adhd command. Once invoked, the prompt modification persists for the entire session, meaning every subsequent response follows the ADHD-optimized format. To revert to standard output, users must explicitly disable the mode by stating "stop adhd mode" or "normal mode."

Implementation Examples

The following examples demonstrate the skill's configuration and resulting prompt structure.

OpenAI Agent Configuration

interface:
  display_name: "I Have ADHD"
  short_description: "Action-first output for ADHD readers"
  default_prompt: "Use $i-have-adhd to make this response action-first and easy to execute."

Resulting System Prompt After Expansion

You are an assistant. <…>  /* original system directives */
[ADHD RULES]
- Lead with the next action.
- Number multi-step tasks.
- End with one concrete next action.
- Suppress tangents…
(Full rule list copied from SKILL.md)

User Activation Command

/i-have-adhd

Summary

• The i-have-adhd skill modifies system prompts by expanding the $i-have-adhd placeholder with rules from SKILL.md.
• Configuration resides in skills/i-have-adhd/agents/openai.yaml and skills/i-have-adhd/agents/gemini.toml.
• Activation occurs via the /i-have-adhd command and persists until "stop adhd mode" is requested.
• The resulting output follows strict formatting: action-first, numbered steps, concrete time estimates, and suppressed tangents.

Frequently Asked Questions

What file contains the ADHD formatting rules in the i-have-adhd repository?

The complete rule set lives in skills/i-have-adhd/SKILL.md. This file contains all behavioral constraints that enforce action-first formatting, numbered steps, and tangent suppression according to the source code.

How do I activate the ADHD-friendly mode when using this skill?

Trigger the skill by entering the /i-have-adhd command. This signals the system to expand the $i-have-adhd placeholder in the current session's system prompt, immediately applying the formatting rules.

Does the ADHD mode persist across multiple messages?

Yes. Once activated via the command, the modified system prompt remains active for the entire session. You must explicitly disable it by saying "stop adhd mode" or "normal mode" to return to standard output formatting.

Which agent configuration files support the i-have-adhd placeholder?

The repository includes skills/i-have-adhd/agents/openai.yaml for OpenAI models and skills/i-have-adhd/agents/gemini.toml for Gemini models, both implementing the $i-have-adhd placeholder for cross-platform consistency.

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