How to Set Up a Development Environment for i-have-adhd: Complete Setup Guide

To set up a development environment for i-have-adhd, clone the repository from GitHub, install a supported AI coding agent (Claude Code, Codex, Cursor, or others), and run the Python test suite in tests/test_run_evals.py to validate the skill's evaluation logic.

The i-have-adhd repository is a specialized skill package for LLM-powered coding assistants that standardizes ADHD-friendly output formatting. Unlike traditional software projects containing application logic, this repository distributes behavioral rules through SKILL.md files that agents load to modify their response style. Setting up a development environment for i-have-adhd requires configuring a local workspace for editing these skill definitions and installing a compatible agent to test behavioral changes in real-time.

Step 1: Clone the Repository

Begin by creating a local copy of the skill package. The repository root contains essential documentation including README.md and the core skill definition at skills/i-have-adhd/SKILL.md.

git clone https://github.com/ayghri/i-have-adhd
cd i-have-adhd

The SKILL.md file encodes the output-style rules that agents follow, while INSTALL.md provides platform-specific installation commands for each supported agent.

Step 2: Install a Supported AI Agent

The i-have-adhd skill is compatible with numerous LLM-powered coding assistants including Claude Code, Codex, Gemini, Zed, Hermes, Pi, Cursor, and OpenCode. According to the ayghri/i-have-adhd source code, each agent requires specific installation commands found in INSTALL.md.

Claude Code:

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

Codex:

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

Cursor or OpenCode:

npx skills add ayghri/i-have-adhd -a cursor -y

# For OpenCode, replace 'cursor' with 'opencode'

Zed: In the Agent Panel, open the Skills manager, select "Create skill from URL", and paste https://github.com/ayghri/i-have-adhd/blob/main/skills/i-have-adhd/SKILL.md.

Gemini CLI:

mkdir -p ~/.gemini/commands
curl -fsSL https://raw.githubusercontent.com/ayghri/i-have-adhd/main/skills/i-have-adhd/agents/gemini.toml -o ~/.gemini/commands/i-have-adhd.toml

Hermes:

hermes skills install ayghri/i-have-adhd/skills/i-have-adhd

After installation, invoke the skill using your agent's specific syntax (e.g., /i-have-adhd for Claude Code, $i-have-adhd for Codex). The skill remains active until you issue "stop adhd mode" as documented in the Persistence section of SKILL.md.

Step 3: Run the Python Test Suite

The repository includes a validation suite in tests/test_run_evals.py that verifies the integrity of the evaluation case catalogue and release-gate logic. This requires Python 3 and either the built-in unittest framework or pytest.

Run the tests from the repository root:

python -m unittest tests/test_run_evals.py

Or using pytest:

pytest tests/test_run_evals.py

The test suite validates three critical components as implemented in ayghri/i-have-adhd:

• test_case_catalog_is_valid_and_balanced: Confirms evals/cases.jsonl contains at least 12 entries spanning 8 or more categories.
• test_score_summary_applies_weights_and_release_gates: Validates that weighted scores calculate correctly according to the release-gate conditions.
• test_candidate_blocker_fails_release_gate: Ensures that blocker rows correctly trigger release-gate failures.

The tests/test_run_evals.py file imports helper functions from scripts/run_evals.py to load and validate the JSON Lines case catalogue stored in evals/cases.jsonl.

Step 4: Edit and Reload the Skill

Development workflows center on editing skills/i-have-adhd/SKILL.md to refine output rules or add new behavioral constraints. After modifying the skill definition:

1. Save changes to skills/i-have-adhd/SKILL.md.
2. Reload the skill in your agent:
   • Claude Code: Restart the session or touch the always-on flag file.
   • Codex: Re-run codex plugin add i-have-adhd@i-have-adhd.
   • Cursor/OpenCode: Copy the updated skill folder to the appropriate directory and restart the agent.

Verify changes by invoking the skill and confirming the output follows updated rules such as "lead-with-action" formatting and numbered steps defined in the Rules section of SKILL.md.

Step 5: Enable Always-On Mode (Optional)

To apply the skill automatically to every session without manual invocation, configure the always-on flag for your specific agent as detailed in INSTALL.md.

For Claude Code:

touch ~/.claude/.i-have-adhd-always

For Cursor or OpenCode, add the markdown rule block to the agent's persistent rules file (e.g., ~/.cursor/rules/alwaysApply.md or ~/.config/opencode/AGENTS.md). The exact markdown snippets are provided in the per-agent sections of INSTALL.md.

Summary

Setting up a development environment for i-have-adhd requires a different approach than traditional software projects:

• Clone the repository to access SKILL.md, INSTALL.md, and the test suite.
• Install one or more supported agents (Claude Code, Codex, Cursor, etc.) using the specific commands in INSTALL.md.
• Run python -m unittest tests/test_run_evals.py to validate the evaluation catalogue and release-gate logic.
• Edit skills/i-have-adhd/SKILL.md to modify behavioral rules, then reload the agent to test changes.
• Optionally configure always-on mode by creating flag files or adding persistent rules.

Frequently Asked Questions

What files should I edit to modify the i-have-adhd skill behavior?

Edit skills/i-have-adhd/SKILL.md to change the output-style rules, formatting constraints, or behavioral instructions that agents follow. This file encodes the core rule set that defines ADHD-friendly formatting such as lead-with-action syntax and numbered steps. Supporting documentation can be updated in README.md or INSTALL.md as needed.

Why does the repository require Python if it contains no application code?

The Python test suite in tests/test_run_evals.py validates the integrity of evals/cases.jsonl and verifies that the release-gate logic in scripts/run_evals.py functions correctly. While the skill itself consists of markdown instructions, the test harness ensures the evaluation catalogue maintains at least 12 entries across 8 categories and that scoring weights apply properly.

How do I test changes without reinstalling the agent constantly?

Most agents support hot-reloading or session-based skill reloading. For Claude Code, touching the always-on flag restarts the skill context. For Codex, running codex plugin remove i-have-adhd followed by codex plugin add i-have-adhd@i-have-adhd refreshes the skill. Directory-based agents like Cursor monitor the skill folder for changes automatically when files are updated.

Can I use i-have-adhd with multiple agents simultaneously?

Yes. The repository supports Claude Code, Codex, Gemini, Zed, Hermes, Cursor, OpenCode, and others. You can install the skill across multiple environments by following the per-agent instructions in INSTALL.md. Each installation references the same SKILL.md rule set, ensuring consistent ADHD-friendly output formatting regardless of which agent you activate.

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