How the Pi / OMP Extension Implements I-Have-ADHD Mode

The Pi / OMP extension implements I-Have-ADHD mode by dynamically injecting ADHD-specific rules from SKILL.md into the model context when enabled, persisting state across sessions via custom entries, and providing CLI flags and slash commands for runtime toggling.

The ayghri/i-have-adhd repository provides a specialized focus mode for neurodivergent users within the Pi and OMP (Open-Multi-Platform) runtimes. This extension seamlessly integrates with the conversation flow by loading rules from skills/i-have-adhd/SKILL.md, managing context injection through the ExtensionAPI, and maintaining state persistence across session lifecycle events.

Rule Loading and Configuration Parsing

The extension initializes by loading two critical resources during startup.

loadRules() reads skills/i-have-adhd/SKILL.md, strips YAML front-matter, and stores the raw rule text in memory for later injection. This function ensures the model receives only the actionable guidance without metadata headers.

loadConfig() parses the optional i-have-adhd.json file for boolean flags including alwaysOn (which enables the mode by default) and hideStatus (which suppresses the UI indicator).

User Interface Controls

The extension registers dual interaction mechanisms for different user workflows.

registerFlag() adds the --adhd CLI flag to the Pi/OMP runtime. When present at startup, this flag forces the mode enabled before any user input processes.

registerCommand() implements the /i-have-adhd slash command, accepting on, off, or no arguments to toggle state at runtime. This command provides immediate feedback through the chat interface regardless of the initial CLI configuration.

Session State Persistence

State survival across conversation tree rebuilds depends on custom session management.

The extension declares custom session entries using the STATE_ENTRY_TYPE constant to store enabled/disabled boolean values. During lifecycle events session_start and session_tree, the getSavedState() function retrieves these entries and restores the previous configuration. The system also listens to session_compact events to maintain synchronization during context compression.

Context Injection and Removal

The syncContext() function in extensions/i-have-adhd.ts manages model context through conditional message injection.

When enabling the mode, the extension sends a RULES_MESSAGE_TYPE custom message containing the rule header and processed SKILL.md content, but only if rulesAreInContext() returns false. This helper function relies on latestMarkerIsActive() from extensions/context-compat.ts to detect whether rules currently exist in the conversation history.

When disabling the mode, the extension injects a DISABLED_MESSAGE_TYPE message containing an "ADHD MODE OFF" notice. This marker signals the model to stop applying ADHD-specific formatting without removing the conversation history.

Input Handling and Stop Phrases

The extension processes every user input through dedicated handlers to detect conversational control signals.

The input handler recognizes the built-in skill alias /skill:i-have-adhd for manual activation. It also monitors for stop phrases including "stop adhd mode" and "normal mode" to trigger automatic deactivation. When detected, the system optionally returns a DISABLE_CONFIRMATION message to confirm the mode change.

When active, the extension displays a green "● ADHD ON" badge via ctx.ui.setStatus() unless the hideStatus configuration flag is set to true.

Practical Usage Examples

Enable ADHD mode at startup using the CLI flag:

pi --adhd

Toggle mode at runtime within a conversation:

await pi.runCommand("/i-have-adhd");

Check active state programmatically from another extension:

if (ctx.sessionManager.hasCustomEntry("i-have-adhd-state")) {
  // Apply ADHD-friendly output formatting
}

Disable mode using natural language input:

await pi.sendMessage({ text: "stop adhd mode" });

Summary

• Rule Management: The extension loads ADHD guidelines from skills/i-have-adhd/SKILL.md via loadRules() and optional settings from i-have-adhd.json via loadConfig()
• State Persistence: Session state survives restarts through custom entries (STATE_ENTRY_TYPE) retrieved by getSavedState() during session_start and session_tree events
• User Controls: The --adhd CLI flag and /i-have-adhd slash command provide flexible activation mechanisms via registerFlag() and registerCommand()
• Context Synchronization: syncContext() injects RULES_MESSAGE_TYPE when enabled and DISABLED_MESSAGE_TYPE when disabled, using rulesAreInContext() and latestMarkerIsActive() for detection
• Conversational Control: Input handlers detect stop phrases like "stop adhd mode" and return DISABLE_CONFIRMATION when turning off the mode
• Visual Feedback: The ctx.ui.setStatus() method displays a green "● ADHD ON" badge unless disabled via configuration

Frequently Asked Questions

How does the Pi extension remember ADHD mode between sessions?

The extension stores the current state in a custom session entry defined by the STATE_ENTRY_TYPE constant according to the ayghri/i-have-adhd source code. When the Pi runtime triggers session_start or session_tree events, the getSavedState() function retrieves this entry from the session manager and restores the previous configuration, ensuring the mode persists across conversation tree rebuilds and application restarts.

Can I hide the ADHD status indicator in the UI?

Yes. Create an i-have-adhd.json configuration file in your project root and set the boolean "hideStatus": true. When this flag is present, the extension skips the ctx.ui.setStatus() call that normally displays the green "● ADHD ON" badge, while maintaining all other ADHD mode functionality including rule injection and context management.

What happens to the model context when I disable ADHD mode?

When disabled through the /i-have-adhd off command or stop phrases, the extension checks if RULES_MESSAGE_TYPE messages remain in the context using rulesAreInContext(). If detected, it injects a DISABLED_MESSAGE_TYPE custom message containing an "ADHD MODE OFF" notice. This marker ensures the model stops applying ADHD-specific rules without surgically removing previous conversation history containing rule applications.

Is this implementation compatible with both Pi and OMP runtimes?

Yes. Both runtimes share the identical implementation in extensions/i-have-adhd.ts because they conform to the same ExtensionAPI contract. The .opencode/plugins/i-have-adhd.mjs file serves as the OpenCode-specific entry point that loads the same TypeScript extension, ensuring consistent behavior across Pi, OMP, and OpenCode platforms.

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