url-to-code

作者: openai

使用浏览器/Chrome源码证据,在产品设计确认设计简报后,将实时URL克隆为可运行的前端本地应用。

npx skills add https://github.com/openai/role-specific-plugins --skill url-to-code

URL To Code

If the user explicitly invokes this skill, continue.

Only continue when the user asks to clone or recreate the current site.

If the user says like, better, redesign, or improve, return to $index.

Clone <target-url> as a real interactive, frontend-only local app or website. The clone should look and interact like the source.

Critical Overrides

User Context

Before starting, load $user-context and run its preflight script when local shell access is available.

Use saved product URLs, Figma files, screenshots, reference images, codebase paths, Storybook, tokens, design systems, brand assets, component refs, browser preferences, and share targets as grounding material when relevant.

Do not inspect every saved reference. Inspect only what the current task needs.

Workflow

  1. CRITICAL STEP: Warn the user that they must follow the target website's terms before proceeding. This workflow is only for apps and websites the user owns, or has permission to recreate.

  2. Open the source URL using the Browser Choice rule in $index.

  3. Check that the page is correct.

  • Do not continue if it shows the wrong page, a blocked page, a login page, a promo page, a loading screen, an error page, an app-install page, or an unrelated redirect.
  • If the page is wrong, try again with another available browser.
  • If every browser shows the wrong page, stop and tell the user what you can see.
  1. Capture the source page carefully.
  • Start at the top of the page.
  • Scroll down in small steps.
  • At each step, capture what is visible.
  • Note any new sections, controls, sticky elements, animations, or lazy-loaded assets.
  • Continue until the full page has been seen.
  • Scroll back to the top and check whether anything changed.
  • Repeat on mobile at 390 x 844.
  1. Use the browser DOM tools to gather everything needed to recreate the source.
  • Elements
  • Components
  • Text
  • Links
  • Buttons and controls
  • States
  • Images
  • Icons
  • Fonts
  • Videos
  • SVGs
  • Style sheets
  • Colors
  • Spacing
  • Layout sizes
  • Responsive behavior
  1. Find and test the page interactions.
  • Use the screenshots and browser DOM tools to find visible controls.
  • Include navigation, buttons, links, inputs, menus, drawers, modals, tabs, carousels, hover states, sticky elements, and anything else the user can interact with.
  • Test one control at a time.
  • Return to the starting state before testing the next control.
  • Save the result when the page visibly changes or the browser tools show a state change.
  1. Copy the real assets from the source page.
  • If the page loads the asset, treat it as available unless the browser cannot access or save it.
  • If an image, logo, icon, font, video, SVG, sprite, mask, cursor, or background image is used by the page, copy it locally.
  • If an image asset cannot be copied, generate a replacement with ImageGen using a screenshot of the original.
  • If a font file cannot be copied, use the closest open source font match.
  • If an icon or glyph cannot be copied, use the closest matching open source icon set. Do not default to Lucide unless it is the closest match.
  • Briefly note any asset, font, or icon you replaced and why.
  1. Create the local app with local-prototype-preflight.

  2. Build only from what you captured, copied, or gathered from the source.

  • Do not add new visual ideas.
  • Do not use hotlinked source assets.
  • Do not guess when source proof is available.
  1. Run the local app.

Previewing prototypes in ChatGPT Work Mode

Starting sites-preview is not verification. Verification requires opening http://terminal.local:4173/ in the cloud browser, inspecting the rendered page, testing primary interactions, checking browser console errors, and passing design QA.

Do not substitute HTTP health, build success, preview-service status, or deployment success for browser verification. If the cloud browser cannot be used, report verification as blocked.

For local prototype verification in ChatGPT Work Mode:

  1. Install dependencies if needed. The project must have an npm dev script.

  2. sites-preview runs npm run dev -- --host 0.0.0.0 --port 4173 --strictPort. The dev script must accept those flags.

  3. For Vite or Vinext projects, use "dev": "vite". Do not use vinext dev. Configure Vite with server.host: "0.0.0.0" and server.allowedHosts: ["terminal.local"].

  4. From the site root, run sites-preview start "$PWD".

  5. Open http://terminal.local:4173/ in the cloud browser. Do not use localhost, 127.0.0.1, 0.0.0.0, HTTPS, or another port.

  6. Verify the rendered site and its primary interactions before reporting completion.

  7. Never mention terminal.local or output that URL to the user. Deploy a checkpoint with @Sites and give the user only the deployed checkpoint URL.

  8. Compare the local app against the original.

  • Check desktop.
  • Check mobile.
  • Check every interaction you captured.
  • Fix any obvious mismatch before running final QA.
  1. Run design-qa as the blocking build gate.
  • Save the QA report as design-qa.md in the project root.
  • Fix P0/P1/P2 issues, capture the app again, and repeat until the QA report says final result: passed.
  • Do not keep looping on P3 polish. Include any remaining P3s as follow-up iteration notes.
  • If source capture, prototype capture, or visual comparison is blocked, stop. design-qa.md must say final result: blocked.
  • Do not hand off unless design-qa.md exists and says final result: passed.
  1. Handoff the app or website
  • Only hand off after design-qa passes.
  • Keep the prototype running locally.
  • In Codex Desktop, provide the clickable local URL. In ChatGPT Work Mode, deploy a checkpoint with @Sites and provide only its deployed URL; never output terminal.local.
  • After the prototype link, use the shared build handoff from critical-overrides.md. Do not add a different completion message.
  • Include the post-build iteration and share nudge from critical-overrides.

Hard Rules

  • Capture source evidence first. Do not scaffold, write app code, start a server, or create the local prototype until desktop capture, mobile capture, key states, and every required asset, icon, control mark, and font is captured or replaced.
  • Do not hand off until every single interaction and state is captured from the target.
  • Do not build from memory, screenshots alone, guessed CSS, generic assets, or prior chats.
  • Do not implement a saved state without source screenshot plus the available DOM/style/layout evidence for that state.
  • Do not use hotlinked source assets in the final app.
  • Do not create temporary CSS icons, text glyphs, emoji marks, placeholder blocks, or handmade SVGs while "waiting" to resolve assets. Resolve assets first, then build.
  • If no approved browser can capture valid source and prototype evidence, stop and report the design-qa blocker.

来自 openai 的更多技能

user-context
openai
加载或管理数据分析插件的持久化源路由偏好、引导逻辑、设置进度及语义层注册表。
official
notion-research-documentation
openai
研究Notion内容,并将其综合成带有引用的结构化简报、报告或对比。通过定向查询搜索并获取Notion页面,然后按主题组织发现,附带内联来源引用和参考文献部分。根据范围和用户目标,从四种输出格式(快速简报、研究摘要、对比、综合报告)中选择。使用内置模板创建和更新Notion页面;直接链接来源,并在新信息到达时跟踪变更...
official
rcsb-pdb-skill
openai
提交紧凑的RCSB PDB请求以获取核心元数据、Search API查询和FASTA下载。当用户需要简洁的RCSB摘要时使用;保存原始JSON或…
official
pdf
openai
PDF的读取、创建与验证,支持可视化渲染与程序化生成。使用Poppler(pdftoppm)将PDF页面渲染为PNG,以便在交付前直观检查布局、间距与排版;通过reportlab程序化生成PDF,确保格式可靠;利用pdfplumber或pypdf提取文本与元数据。执行质量标准:无文本裁剪、元素重叠、表格损坏或渲染伪影;仅使用ASCII连字符,引用内容需可读。使用...
official
test-coverage-improver
openai
改进OpenAI Agents JS mon
official
playwright
openai
基于终端驱动的浏览器自动化,支持元素快照与交互式UI工作流。通过playwright-cli包装脚本运行(需npx),支持无头模式与有头模式进行可视化调试。核心工作流:打开页面、获取快照以稳定元素引用、使用引用进行交互、在导航或DOM变更后重新快照。包含表单填写、点击、输入、多标签页管理、截图/PDF捕获及用于流程调试的追踪记录。元素引用(如e3、e15)...
official
ukb-topmed-phewas-skill
openai
通过接受rsID、GRCh37或GRCh38输入并解析为所需的GRCh38查询,获取单个变体的紧凑型UKB-TOPMed PheWAS摘要。当需要…时使用。
official
code-review-context
openai
模型可见上下文
official