AI 工作流工具

AI 编程上线检查清单

为 AI 生成应用和 AI 辅助改动生成本地上线检查清单:构建、环境变量、登录、支付、SEO、移动端、文档与发布记录。

本地 AI 工作流工具

AI 编程上线检查清单

为 AI 生成代码生成上线检查清单。不上传内容;勾选状态和备注只保存在当前浏览器。

进度0/15
Must0/7
Should0/7
Optional0/1
Cursor / editor agent · Web app / landing page · Public website

Open the changed files in Cursor, Copilot, Claude Code, Codex, Windsurf, or your git client and confirm the agent only touched the intended scope.

Run the command that your host will run, such as npm run build, next build, mvn package, flutter build, or the platform build step. A dev server is not enough.

Open a fresh browser profile or incognito window and complete the key path as a real user: landing, sign in if needed, create or submit data, and finish the expected action.

Search for secrets, service-role keys, private tokens, test credentials, and variables accidentally exposed to the frontend after AI edits.

Confirm build command, output directory, runtime version, region, environment variables, custom domain, CORS, redirects, and callback URLs.

Search visible pages, README, errors, empty states, and meta tags for lorem ipsum, TODO, fake testimonials, model disclaimers, or generic marketing text.

Check 360px to 430px widths for horizontal scroll, fixed buttons covering inputs, clipped dialogs, overflowing cards, and long words breaking layout.

Force failed network calls, empty lists, invalid forms, expired sessions, and 404 or not-found states. AI-built happy paths often miss these.

Run the checks that exist in the project. If there are no automated tests, write a short manual test list instead of pretending coverage exists.

Document install, local dev, build, deploy, required env vars, known limitations, and rollback steps so the next AI session or teammate has context.

For payments or external services, confirm repeated webhook delivery does not create duplicate orders, credits, emails, or records.

For public pages, inspect meta title, description, canonical URL, robots rules, sitemap inclusion, Open Graph, and whether placeholder pages are indexable.

Make sure headings, FAQ answers, definitions, and tool descriptions give short, factual answers that search engines and AI assistants can quote accurately.

For APIs or external integrations, verify request payload shape, status codes, auth headers, and cURL examples that AI may have invented.

Tab through the main UI, verify focus is visible, labels are attached to inputs, buttons have names, and dialogs can be closed without a mouse.

本工具属于「AI 工作流工具」工作流。 查看完整工作流 · AI 编程项目上线前检查清单

关于此工具

AI 编程上线检查清单是一个本地 launch checklist,面向 AI 生成应用和 AI 辅助代码改动。它适合使用 Cursor、GitHub Copilot、Claude Code、Codex、ChatGPT、Windsurf、Replit Agent、Bolt、Lovable 等工具之后,继续检查模型写完代码后仍然必须完成的交付动作。

如何使用

选择 AI 编程来源、项目类型、技术栈、上线目标和高风险功能。工具会生成一份可勾选清单,用于发布前检查,或作为下一轮 AI agent 的上下文。

  1. 选择代码来源:编辑器 agent、Copilot、CLI agent、复制 ChatGPT 代码、vibe coding builder,或人工和 AI 混合修改。
  2. 选择项目类型、技术栈、上线目标,以及登录、支付、数据库、公开 SEO、多语言、文件上传、管理后台、外部 API 等功能。
  3. 按 Must、Should、Optional 完成检查,并记录 build 输出、env 审查、移动端测试、SEO 检查和发布说明等证据。
  4. 复制 Markdown 报告,放进 PR、issue、README、release note,或下一轮 AI 编程 prompt。

功能特点

首版不上传代码、不调用大模型,也不声称自动判断代码质量。它只根据你的选择,在本地生成可执行检查清单。

  • 覆盖 AI 生成 Web 应用、API、移动端发布、浏览器工具和 vibe coding 原型。
  • 检查构建命令、环境变量、密钥、登录、支付、webhook、SEO、移动端布局、文档、日志和回滚。
  • 在浏览器本地保存配置、勾选状态和证据备注。
  • 导出 Markdown,用于 PR、问题跟踪、发布记录和 AI agent 交接 prompt。

常见问题

什么是 AI coding checklist?

AI coding checklist 是给 AI 生成或 AI 修改代码使用的发布检查清单,重点是验证:diff 审查、构建命令、测试、密钥、部署配置、登录、支付、SEO、移动端布局和发布记录。

这个工具会扫描我的代码仓库吗?

不会。清单只根据你选择的选项在本地生成,不上传文件、不读取仓库、不调用 AI 模型,也不会自动判断代码质量。

这个工具适合谁?

适合使用 Cursor、Copilot、Claude Code、Codex、Windsurf 的开发者,也适合用 Bolt、Lovable、Replit Agent 或 ChatGPT 生成代码并准备上线的新手用户。

vibe coding 项目能用吗?

可以。选择 AI-built prototype 或 vibe coding builder 相关选项后,清单会包含原型边界、占位文案、环境变量、移动端检查和公开发布风险。