用于约束AI在“胶水开发”模式下的行为,要求优先、直接、完整复用成熟仓库与生产级库,禁止复制、裁剪、重写或降级封装依赖,禁止Mock、Stub、占位实现,只生成最小必要的胶水代码与业务层调度逻辑。
中文版提示词
以下是胶水开发约束,请严格遵守:1.不得自行实现底层或通用逻辑,必须优先、直接、完整复用既有成熟仓库与生产级库;2.不得为了方便而复制依赖库代码到当前项目中再修改使用;3.不得对依赖库进行任何形式的功能裁剪、逻辑重写或降级封装;4.允许使用本地源码直连或包管理器安装方式,但实际加载的必须是完整生产级实现;5.不得使用简化版、替代版或重写版依赖冒充真实库实现;6.所有依赖路径必须真实存在并指向完整仓库源码;7.不得通过路径遮蔽、重名模块或隐式fallback加载非目标实现;8.代码中必须直接导入完整依赖模块,不得进行子集封装或二次抽象;9.不得在当前项目中实现依赖库已提供的同类功能;10.所有被调用能力必须来自依赖库的真实实现,不得使用Mock、Stub或Demo代码;11.不得存在占位实现、空逻辑或“先写接口后补实现”的情况;12.当前项目仅允许承担业务流程编排、模块组合调度、参数配置与输入输出适配职责;13.不得在当前项目中重复实现算法、数据结构或复杂核心逻辑;14.不得将依赖库中的复杂逻辑拆出后自行实现;15.所有导入的模块必须在运行期真实参与执行;16.不得存在“只导入不用”的伪集成行为;17.必须确保sys.path或依赖注入链路加载的是目标生产级本地库;18.不得因路径配置错误导致加载到裁剪版、测试版或简化实现;19.在生成代码时必须明确标注哪些功能来自外部依赖;20.在任何情况下不得生成或补写依赖库内部实现代码;21.只允许生成最小必要的胶水代码与业务层调度逻辑;22.必须假设依赖库为权威且不可修改的黑箱实现;23.项目评价标准以是否正确、完整站在成熟系统之上构建为唯一依据,而非代码量。
英文版提示词
The following are glue-code development constraints that must be strictly followed: 1. Do not implement underlying or generic logic yourself; prioritize, directly, and fully reuse mature existing repositories and production-grade libraries. 2. Do not copy dependency-library code into the current project and then modify it for convenience. 3. Do not perform any functional trimming, logic rewriting, or downgraded wrapping of dependency libraries. 4. Local source direct-linking or package-manager installation is allowed, but what is actually loaded must be a complete production-grade implementation. 5. Do not use simplified, alternative, or rewritten dependencies to impersonate real library implementations. 6. All dependency paths must genuinely exist and point to the complete repository source code. 7. Do not load non-target implementations via path shadowing, duplicate module names, or implicit fallbacks. 8. Code must directly import complete dependency modules; no subset wrapping or secondary abstraction. 9. Do not implement in the current project functionality already provided by the dependency library. 10. All invoked capabilities must come from the real implementation of the dependency library; no Mock, Stub, or Demo code. 11. No placeholder implementations, empty logic, or "interface first, implement later" situations. 12. The current project may only handle business-flow orchestration, module composition/scheduling, parameter configuration, and input/output adaptation. 13. Do not re-implement algorithms, data structures, or complex core logic in the current project. 14. Do not extract complex logic from the dependency library and implement it yourself. 15. All imported modules must genuinely participate in execution at runtime. 16. No "import but never use" pseudo-integration. 17. Ensure sys.path or the dependency-injection chain loads the targeted production-grade local library. 18. Do not load trimmed, test, or simplified implementations due to path-configuration errors. 19. Clearly annotate which functionality comes from external dependencies when generating code. 20. Never generate or backfill dependency-library internal implementation code. 21. Only generate the minimal necessary glue code and business-layer orchestration logic. 22. Assume the dependency library is an authoritative, unmodifiable black-box implementation. 23. The project is evaluated solely on whether it is correctly and completely built on top of mature systems, not on the amount of code. 🛠️ **适用 AI 工具**:Claude、ChatGPT、Cursor、Copilot、DeepSeek、Gemini

◯ 评论 0