About this Project
This project provides a production-quality, reusable skill framework designed for AI coding agents. Its primary goal is to enable these agents to contribute effectively to mature open-source repositories, mimicking the behavior of experienced senior engineers. It's not just a single prompt but a structured engineering framework that guides an AI agent to:
- Understand a repository before making changes.
- Keep code modifications minimal.
- Adhere to local coding conventions.
- Place new logic within the correct architectural layer.
- Write tests in the repository's native style.
- Prepare commits and pull requests that are maintainer-friendly.
- Document engineering knowledge gained from each contribution.
Key Features
- Comprehensive Framework: Includes detailed guides, checklists, and templates for various stages of the development process, from investigation to bug fixing and feature implementation.
- Repository Adaptation: Emphasizes understanding and adhering to existing repository conventions, style, and architecture.
- Maintainer-Friendly Output: Focuses on producing minimal diffs, correct architectural placement, and well-documented, review-ready pull requests.
- Agent Agnostic: Designed to be portable and work with various AI coding agents that support reusable prompt folders or skills.
- Structured Skills: Organizes functionality into distinct skills like senior-open-source-engineer and an alias seo for easier invocation.
- Detailed Guidance: Provides specific rules and prompts for different phases of development, including investigation, implementation, testing, and review.
Supported Tools
This framework is intended for use with AI coding agents such as:
- Codex CLI
- Claude Code
- Gemini CLI
- Cursor
- Roo Code
- Cline
- Amp
- And other similar AI coding assistants.
Installation and Usage
The framework is distributed as plain folders with SKILL.md entrypoints, ensuring portability. Users can copy the provided skill folders into their agent's skill or prompt directory. Specific installation and usage instructions are provided for various agents like Codex CLI, Claude Code, Cursor, Gemini CLI, and others, detailing how to integrate and invoke the skills effectively.
Contributing
Contributions to this framework should maintain its core design principles: prioritizing repository adaptation, minimal diffs, and maintainability over sheer code volume. Improvements should enhance the skill's repository-nativeness, portability, or ease of maintenance.