Agent Ops Create Python Project

Skill: agent-ops-create-python-project

Create a plan and issues for implementation of a production-ready Python project

Templates: See TEMPLATES.md for pyproject.toml, build.py, README, AGENTS.md, .gitignore

Triggers

• User asks to create a new Python project
• User provides project requirements/discussion to convert into implementation
• User wants to scaffold a CLI tool, library, or application in Python

Procedure

Phase 1: Requirements Gathering

1. If input provided: Analyze the discussion/requirements
2. If no input: Interview user about:
   • Project name and purpose
   • Core features/commands
   • External services/APIs needed
   • Data processed (files, APIs, databases)
   • CLI, library, or both?
   • Specific dependencies?

Phase 2: Architecture Design

Extract from requirements:

• Features: List all behaviors/capabilities
• Data Models: Entities and relationships
• Dependencies: Map features to PyPI packages
• Modules: Cohesive, loosely-coupled units
• Interfaces: Public APIs per module

Phase 3: Issue Creation

Create issues for:

1. Project scaffold (pyproject.toml, README, .gitignore, AGENTS.md)
2. Build pipeline (scripts/build.py)
3. Configuration (src/package/config.py)
4. CLI layer (src/package/cli.py)
5. Core modules (one issue per module)
6. Test infrastructure (tests/conftest.py)
7. Unit tests (tests/unit/*)

Phase 4: Plan Generation

Create plan with:

• Dependency order (scaffold → config → core → CLI → tests)
• Estimated effort per issue
• Quality gates between phases

Project Structure

project-name/
├── pyproject.toml          # Config, dependencies, tools
├── README.md               # Overview, install, usage
├── AGENTS.md               # AI agent guidelines
├── .gitignore
├── scripts/
│   └── build.py            # Build pipeline
├── src/<package>/
│   ├── __init__.py
│   ├── cli.py              # Thin CLI (typer)
│   ├── config.py           # Configuration
│   └── <modules>.py        # Core logic
└── tests/
    ├── conftest.py
    ├── unit/
    └── integration/

Code Standards

Design Principles

• SRP: One responsibility per module/function
• DRY: No duplicated logic
• Dependency Injection: Accept dependencies as parameters

Architecture Rules

• Thin CLI: Parse/format only, delegate to core
• No src imports: from <package> not from src.<package>
• Config via environment: .env + python-dotenv

Function Guidelines

• Max 15 lines per function
• Max 3 levels of nesting
• Type annotations on ALL functions

Output Checklist

• All functions have type annotations
• Tests cover core logic (≥75% target)
• CLI is thin orchestration layer
• pyproject.toml is valid
• README explains uv run usage
• AGENTS.md contains guidelines