Metadata-Version: 2.4
Name: agent-tools-drnt
Version: 0.1.5
Summary: Worktree-based agent workflow management toolkit
Author: Agent Tools Contributors
License: MIT
Requires-Python: >=3.11
Requires-Dist: colorama>=0.4.6
Provides-Extra: cli
Requires-Dist: colorama>=0.4.6; extra == 'cli'
Provides-Extra: dev
Requires-Dist: pytest-cov>=4.0.0; extra == 'dev'
Requires-Dist: pytest-xdist>=3.0.0; extra == 'dev'
Requires-Dist: pytest>=7.0.0; extra == 'dev'
Requires-Dist: ruff>=0.1.0; extra == 'dev'
Description-Content-Type: text/markdown

# agent-tools

Worktree-based agent workflow management toolkit.

## Overview

`agent-tools` provides a Python-based CLI for managing isolated Git worktrees for agent workflows. It replaces shell scripts with a maintainable, testable Python package that can be used locally, as a submodule, or in CI/CD pipelines.

## Features

- **Isolated Worktrees**: Each agent gets its own isolated Git worktree
- **Simple CLI**: `start`, `run`, `commit`, `push`, `clean` commands
- **Manual PR Creation**: Push branch and create PR manually in GitHub UI
- **Cross-platform**: Works on Windows, Linux, and macOS
- **Installable**: Can be installed via pip or used as submodule

## Quick Start

### Installation

```bash
# From source
cd agt
uv pip install -e .  # or: pip install -e .

# Or as submodule
git submodule add -b main https://github.com/tnedr/agent-ops .tools
cd .tools/agt
uv pip install -e .  # or: pip install -e .
```

**Note**: This project follows UV Cache Workflow Guidelines. If using `uv`, ensure `UV_CACHE_DIR` is set (e.g., `E:\uv-cache` on Windows).

### Usage

```bash
# Start a new agent worktree
agt start

# Run commands in the worktree
agt run python my_script.py

# Commit changes
agt commit "feat: new feature"

# Push to remote (then open PR manually)
agt push

# Optional: cleanup after PR is merged
agt clean
```

### VS Code Command Runner

The project includes a Command Runner integration for VS Code/Cursor:

1. **Generate Command Runner settings:**
   ```bash
   python scripts/update_command_runner.py
   ```

2. **Use commands in VS Code/Cursor:**
   - Press `F1` → `Command Runner: Run...`
   - Select `math.multiply`, `time.now`, or `env.check`
   - Or use in chat: `> run command: math.multiply 7 6`

3. **Available commands:**
   - `math.multiply`: Multiply two numbers (a×b)
   - `time.now`: Get current UTC timestamp
   - `env.check`: Verify venv and colorama dependency

4. **Agents can use these commands:**
   ```bash
   agt run python ${workspaceFolder}/.tools/scripts/multiply.py 3 7
   agt run python ${workspaceFolder}/.tools/scripts/venv_check.py
   ```

## Documentation

- [Quick Start Guide](docs/quick_start.md)
- [CLI Reference](docs/cli_reference.md)
- [Architecture](docs/architecture.md)
- [Changelog](docs/CHANGELOG.md)

## Examples

See the [examples/](examples/) directory for:
- `demo_local.sh`: Local workflow example
- `demo_action.yml`: GitHub Actions workflow example

## Development

### Setup

```bash
cd agt
uv pip install -e ".[dev]"  # or: pip install -e ".[dev]"
```

**Note**: Uses UV Cache Workflow - minimal `.venv`, global cache.

### Run Tests

```bash
uv run pytest tests/ -v  # or: pytest tests/ -v
```

### Run Linter

```bash
uv run ruff check .  # or: ruff check .
```

## Project Structure

```
agt/
├── __init__.py          # Package metadata
├── __main__.py          # Allows `python -m agt`
├── worktree.py          # Low-level Git worktree operations
├── cli.py               # CLI entrypoint and command handlers
└── pyproject.toml        # Package configuration

bin/
└── agt                  # Standalone entrypoint script

docs/                    # Documentation
tests/                   # Test suite
examples/                # Usage examples
.github/                 # CI/CD workflows and actions
```

## License

MIT

