The agent that writes the code never reviews it. Spec-driven TDD with agent separation, adversarial QA, and dynamic rigor for Claude Code.
推荐理由
README 将它定位为「The agent that writes the code never reviews it」,核心痛点是让 AI agent 通过可复用 skills 和工具流程处理具体任务。它有一定社区验证,同时仍保留发现潜力,license 清晰,主要技术栈是 Shell,适合作为「AI agent 工具链」的候选项目。
注意事项
项目还偏早期,需要重点验证核心路径是否稳定;README 摘要信息有限,发布前建议再人工扫一遍文档;展示前建议跑通 README quickstart,并确认部署成本、外部依赖和数据安全边界。