agent-tools 🛠️

agent-tools is a Python CLI suite designed to help with AI-powered development workflows.

Features

• Jules Integration: Create and manage Jules coding sessions directly from your terminal.
• Agent Prompts: Define specialized AI agents (like bolt for performance) with custom system prompts in a simple YAML configuration.
• Kit: Install reusable prompt commands into AI coding agents (opencode, kiro, gemini) from a shared template library.

Installation

This project uses uv for fast, reliable dependency management.

# Clone the repository
git clone https://github.com/scartill/agent-tools.git
cd agent-tools

# Install dependencies and create a virtual environment
uv sync

Usage

Kit Commands

kit manages prompt command installations for supported AI coding agents. It reads component definitions from the built-in templates/ directory and writes agent-specific command files into your project.

Supported agents: opencode (default), kiro, gemini

kit commands add

Install a prompt component for the selected agent.

uv run agent-tools kit --agent <agent> commands add <component>

Arguments:

• <component>: Name of the template component to install (e.g. sc.superb.critique).

Options:

• --agent: Target AI agent — one of opencode, kiro, gemini (default: opencode).

Examples:

# Install the "superb critique" prompt for opencode (default)
uv run agent-tools kit commands add sc.superb.critique

# Install the same prompt for kiro
uv run agent-tools kit --agent kiro commands add sc.superb.critique

# Install the drift-detection prompt for gemini
uv run agent-tools kit --agent gemini commands add sc.superb.drift.detect

The command writes the transformed prompt file to the agent's conventional location:

Agent	Output path
opencode	.opencode/command/<component>.md
kiro	.kiro/prompts/<component>.md
gemini	.gemini/commands/<component>.toml

Available Components

Component	Compatible agents	Description
sc.superb.critique	opencode, kiro, gemini	Tailored prompt for /speckit.superb.critique
sc.superb.drift.detect	opencode, gemini	Detect drift between manually modified code and existing specification

Jules Commands

Create an automated coding session.

uv run agent-tools jules create \
  --repository owner/repo \
  --branch feature-branch \
  --agent bolt \
  --title "Performance optimization for API"

Options:

• --repository, -r: Target GitHub repository (owner/repo).
• --branch, -b: Branch for Jules to work on.
• --agent, -a: Name of the agent defined in prompts.yaml.
• --title, -t: (Optional) Title for the session.

Configuration

The tool requires API credentials to interact with external services. You can provide these via environment variables, a .env file, or CLI flags.

Credentials

• JULES_API_KEY: Required for interacting with the Jules API.
• GITHUB_PAT: GitHub Personal Access Token for repository interactions.

Example .env file:

JULES_API_KEY=your_jules_key
GITHUB_PAT=your_github_token

Agent Prompts (prompts.yaml)

Define your AI agents in prompts.yaml in the project root:

agents:
  bolt:
    prompt: |
      You are "Bolt" ⚡ - a performance-obsessed agent...
      (Full prompt text here)

Project Structure

• src/agent_tools/cli.py: CLI entry point and command registration.
• src/agent_tools/config.py: Configuration and environment management.
• src/agent_tools/clients/: API clients for Jules and GitHub.
• src/agent_tools/commands/: Implementation logic for CLI commands.
• kit/: Reusable tool definitions and prompts for agent-led workflows.

Development

Running Tests & Linting

# Lint with ruff
uv run ruff check .

# Fix linting issues
uv run ruff check --fix .

Roadmap

• Add jules status to track session progress.
• Integration with GitHub Copilot Review API.

License

MIT License - Copyright (c) 2026 Boris Resnick