oil-oil

Generate UI design mockups and help reconstruct generated UI screenshots into HTML/CSS. Prefer built-in image generation when available; use ZenMux + GPT Image 2 only as fallback or for scripted local outputs. TRIGGER when the user says "生成图片", "画图", "设计 UI", "UI 设计", "出图", "create an image", "design a screen", "landing page", "设计稿还原", "截图还原 HTML", "把图片复刻成网页", or when another skill needs image generation.

Delegate coding and image-generation tasks to Codex CLI for execution. Only invoke this skill when the user explicitly asks to use Codex — e.g., "用 codex 来做", "让 codex 执行", "ask codex to...", "codex 帮我写", "让 codex 生成/画一张图". Do not proactively delegate to Codex for requests the user didn't specifically ask Codex to handle. Codex is an autonomous agent with the same tools as Claude (file read/write, grep, bash) plus a built-in image_gen tool — it explores the codebase and implements changes on its own. Codex is strong at execution and writing code but comparatively weak at understanding ambiguous problems and high-level design, so Claude's job is to do the understanding and design first, then hand Codex a clear, well-scoped task to execute.

Run a structured UI/UX consultation to either (a) co-design a project-specific design system and emit `design-spec.md`, (b) review an existing UI with prioritized fixes, or (c) emit compact do/don't rules for a surface. Triggers when the user wants to define / build / refine a design system or design tokens, asks for a design spec, asks for a full UI review of a screen / mockup / PR, or wants design rules for a surface type. Do NOT trigger for narrow one-off questions ("is this color OK?", "should this button be larger?") — answer those directly without invoking the consultation flow.

Use Grok 4.5 as the required external design advisor when a task needs UI critique, UX critique, component-choice review, interaction-flow review, design imagery markdown, art direction, visual hierarchy judgment, design-system fit, color/type/layout suggestions, HTML mockups, SVG icons, handwritten SVG wordmarks, or file-based feedback on existing UI. Trigger on requests like "design a page", "generate design imagery", "give design advice", "optimize this UI", "improve UX", "optimize user flow", "review interaction", "check usability", "reduce friction", "choose the right component", "review this design", "create an icon", "suggest colors", "UI mockup", "visual direction", "lettering", "wordmark", "handwritten logo", or when another agent needs a second opinion on visual or interaction quality.

Create a bold, visually striking landing page — unconventional layouts, GSAP scroll animations, SVG elements, clip-path dividers, real depth and layering. Use this skill whenever the user wants a landing page, homepage, product page, marketing page, campaign page, event page, or any single-page site that needs to impress. Trigger on: 帮我做一个落地页, 做个首页, 产品展示页, 活动页, 营销页, 官网首页, landing page, product page, promo page, make me a homepage, build a product showcase, create a campaign page. Workflow: asks one question about the product, then opens a live browser preview of 57 city-inspired visual styles to click-select, lets the user pick typography/nav/color tone/hero variant/features variant/page sections interactively, then outputs a complete multi-file site (index.html + style.css + main.js + SVG sprite) with real product copy.

Redesign and polish GitHub repository README homepages around the real project theme. Use when a user asks to beautify, decorate, redesign, rebrand, visually upgrade, or simplify a GitHub README; create SVG hero banners, section headers, showcase walls, before/after modules, or a cohesive repository homepage; or make a README feel like a designed landing page while keeping it readable and maintainable.

AI 英语外教 Lumina——一个住在用户飞书 Base 里、有完整人格与持续记忆的双语学习合伙人。当用户说'学英语'、'练口语'、'lumina'、'飞书外教'、'每日英语'，或想开始/继续一段英语练习时使用。