要解决的问题The problem
很多 AI 生成的 Demo 能运行，却缺少 Git、测试、结构和持续迭代所需的工程基础。Many AI-generated demos run, but lack the Git, tests, structure and iteration habits that sustain an engineering project.
实现思路Approach
通过导入、诊断、白话解释、任务闯关与复盘形成学习闭环，而不是直接替用户重写代码。Import, diagnosis, plain-language explanation, guided tasks and review form a learning loop, instead of rewriting the code for the user.
公开证据Public evidence
- 支持 GitHub、ZIP 与本地项目导入到隔离工作区Imports GitHub, ZIP and local projects into an isolated workspace
- 实现超大文件、重复代码、Git、CI、测试五类工程债诊断Detects five kinds of engineering debt: oversized files, duplicated code, Git, CI and tests
- 包含 Web 主应用、Electron 桌面端、离线兜底与演示样本Includes a web app, an Electron desktop client, an offline fallback and demo samples
当前限制Current limits
- README 明确标记为 MVP 提议版，根目录目前没有测试或 CIThe README marks it as an MVP proposal; there are currently no tests or CI at the repository root
- 桌面层包含注明来源的第三方 vendored 代码，案例只计算自研诊断与学习闭环The desktop layer vendors third-party code with source attribution; this case counts only the self-built diagnostics and learning loop