suyuan2022/ai-dev-pipeline

This repository is a collection of 12 skills for Claude Code, Anthropic's AI coding assistant. Together they form a structured development pipeline meant to take a software idea all the way from initial concept through to a tested, reviewed pull request, with the AI handling most of the work at each stage. The pipeline starts with a skill that interviews you about your idea and pushes for missing details, then moves through creating a product requirements document, splitting the work into development tasks, writing code using test-driven methods, reviewing code quality, running end-to-end verification against acceptance criteria, and fixing bugs if verification fails. There are also skills for manual testing sessions, handing off context when an AI session is running low on memory, and improving code architecture over time. Only three of the twelve skills require active human input: refining the idea at the start, confirming the requirements document before work begins, and participating in manual testing at the end. The rest run with minimal intervention. Installation involves cloning the repository and running a short shell script that creates symbolic links from the skills folder into the directories where Claude Code and Codex look for custom skills. After that, a single git pull updates both tools at once. The README also includes a routing template written for the CLAUDE.md configuration file. Pasting it into your project tells the AI which skill to reach for in each situation, such as starting a new feature, debugging an error, or handling a context-full session. This project is primarily documented in Chinese. It is released under the MIT license.