Image

Generate publication-quality AI illustrations for academic papers using Gemini image generation. Creates architecture diagrams, method illustrations with Claude-supervised iterative refinement loop. Use when user says "生成图表", "画架构图", "AI绘图", "paper illustration", "generate diagram", or needs visual figures for papers.

Claude skill: write a one-page US resume from your real details, with the free CV Chap Chap connector for checks, designs and a finished draft.

Reviews a website and produces a friendly, evidence-based audit that separates what was actually observed from suggested improvements and items that still need testing. Use this skill whenever the user asks to audit, review, critique, or sanity-check a website or landing page, asks what could be improved on a site (clarity, navigation, mobile experience, accessibility basics, trust signals, calls to action, technical hygiene), or wants a UX, conversion, or site-quality review with honest, non-exaggerated findings, even if they don't say "audit".

Write or revise copy-ready prompts for MiniMax H3 Image. Use for text-to-image briefs, posters with exact text, and edits that preserve details from a reference image.

Check whether an AI-generated on-model product photo (a person wearing or holding a product) still shows the real product. Use when the user gives a real product photo and a generated result photo and asks whether it is faithful, ready to publish, or what went wrong. Produces a scorecard with a Verified verdict and short fixes.

Use this skill whenever the user wants to add text graphics, typography, decorative patterns, frames, illustrations, badges, or postcard/poster layouts onto a photograph (such as birthday posts, travel/camping memories, anniversaries, or social media cards). Provides an end-to-end professional Art Director workflow covering visual composition analysis, directional flow, color harmony, 12+ curated design styles, authentic multi-tone Thai/English copywriting, and 4K Retina HTML/CSS/SVG vector rendering without distorting the original photo.

End-to-end App Store and Google Play screenshot sets. Finds the app's benefits in the codebase, captures real screens from iOS simulators and Android emulators (iPhone, iPad, iPhone Duo, Android phone and tablet) in every language, composes each tile on an HTML canvas with device frames, headlines and callouts from locale JSON files, and renders every locale x device with Playwright. Use to create a store screenshot set, add a language, add a device size, or rebuild a set where text and UI must stay exact.

Use when driving Adobe Photoshop with standalone UXP scripts (.psjs): write, launch, poll, and verify document create/edit/export from an agent. Covers executeAsModal, batchPlay, file tokens, doc-id safety, and psjs vs UXP plugins vs ExtendScript/JSX. Chinese triggers: psjs驱动Photoshop, UXP脚本, batchPlay, executeAsModal, 用脚本改PSD, 从命令行跑Photoshop脚本, 图层样式脚本, 导出切片, 切图工作流, 九宫格, 切图清单, 布局JSON. Do NOT use for UXP plugin/panel or MCP-bridge development, ExtendScript/JSX-only work, or read-only PSD inspection that psd-tools/ag-psd can do without launching Photoshop.

Prepare a brief for generating an on-model product photo with an image engine. Use when the user gives a product photo and wants a prompt, the risks to watch, and the resolution to use before they generate. It tags the product, assembles a starter prompt from the recipe rules, and says what to check afterwards. It does not generate images.

Turn a photo or a line about someone's day into a page of a paper pop-up book they can play with, delivered as ONE self-contained HTML file (three.js, no server, no install, works offline). Layers of cut paper fold up from the gutter; the reader taps a piece, drags a pull-tab, turns a sun-and-moon dial between day and night, and something in the scene moves on its own. Use this whenever someone sends a photo, a place, a memory or a one-line diary entry and wants it as a pop-up book, a paper diorama, a "popbook" or a playable keepsake page, or wants to add a page to a popbook they already have, even if they never say "pop-up".

Turn a few photos of yourself (or you and your partner, friend, team) into illustrated WhatsApp sticker packs that still look like you. Guides the whole run, from picking reference photos, choosing a style and an expression list, to writing a generation brief for any image model (ChatGPT, Codex, Gemini) and packing the results into 512×512 WebP files under 100KB with a tray icon and an overview sheet. Use when the user says "make stickers of me", "做我的 sticker", "我的形象表情包", "个人 IP sticker", "WhatsApp sticker pack", "情侣 sticker", or hands over photos and asks for stickers.

Builds a code-generated Open Graph share image (the 1200x630 card that shows up when a link is pasted into Slack, Bluesky, iMessage, LinkedIn, Discord, or X), using the site's real brand colors and fonts and a mock of what the product actually does. Wires up the meta tags for Next.js, Astro, or any static site, then renders the PNG and checks it by eye and with a bundled tag checker. Use whenever the user asks for an "OG image", "share image", "social card", "link preview", "unfurl image", "Twitter card", or says links to their site look bad when shared. Not for favicons, app icons, or general marketing graphics.

独立的非人像摄影后期工作流，面向任何能看图的 agent。用户发来一张或多张照片并要求修图（例如“修”“修一下这张”“修成成片”“精修”）时使用，适用于建筑、街景、风光、园林、夜景、室内、静物、展品等非人像照片。自动读片、写导演单、指挥生成式光色与确定性调色、检查是否符合自然光的物理规律和用户偏好的力度，直接交付一张成片；按宿主实际具备的工具分级执行。

根据用户上传的IP人物参考图和指定年份，策划并生成单月或全年12张3:4竖版IP月历海报，兼顾角色与系列一致性、可选的固定或季节换装、水彩月历卡片，以及准确的公历、农历、节气和重要节日。用于“IP月历”“IP日历图”“IP年份日历”“全年角色月历”等请求；普通无IP日历、纯日期查询或未要求月历的角色插画不自动触发。

Generate or edit raster images with Microsoft Foundry MAI-Image models (default MAI-Image-2.6-Flash) when the task benefits from AI-created bitmap visuals such as photos, illustrations, textures, sprites, mockups, posters, or chroma-key cutouts. Use when the agent should create a brand-new image, edit or combine up to five existing images, or derive visual variants from references, and the output should be a bitmap asset rather than repo-native code or vector. Do not use when the task is better handled by editing existing SVG/vector/code-native assets, extending an established icon or logo system, or building the visual directly in HTML/CSS/canvas.

Produce reference-locked 2D sprite sequences with four-frame reconstruction guides, original-cel repairs, explicit pixel edits, real video timestamps and fixed-canvas exports.

Turn someone's own handwriting into a reusable TrueType font and a stable profile, then lay out any text in that handwriting and export a print-ready PDF. Inputs: a photo of their writing, a practice sheet they fill in, or a PDF that already shows their handwriting (embedded font or text layer). Fully local (no network, no cloud), Windows/macOS/Linux. Use when the user uploads a handwriting image or PDF and asks for a handwritten-style document, letter, note or font; also use to build a personal 'my handwriting' font once and reuse it.

Use when compressing, resizing, or converting bulk product images (PNG, JPG, WEBP, AVIF) to web-ready JPG, WebP, or AVIF files in this repository, setting up a user's working folder on a new machine, or troubleshooting height-only sizing, aspect-ratio errors, and CLI flags.

在 Codex 中按四阶段制作手绘卡通肖像海报，优先用弹出选择栏与填空收集照片、风景、装饰和改稿意见。人物采用粗黑外轮廓、轻笔内线、暖象牙亮面与克制的浅赭阴影，交付独立透明图层和拼接稿。

判断一张图片是 AI 生成、P 过图，还是重复图。三层取证：元数据（AIGC 国标标识 / C2PA / EXIF）、感知哈希查重、可选像素模型。用于电商签收图审核、素材溯源、批量图片筛查等场景。

先对照参考图判断指示牌、钥匙扣等平面图哪些轮廓应去除、哪些应保留，再调用本地 Photoshop 按区域将多余外框抠成透明并逐图验收。保留实体牌面与图案，自动复判不确定区域并汇总异常，不等待人工实时圈选。

Make 8-bit / pixel-art versions of fire performance sprites with rebuilt fire. Use after the fire-sprites skill has rendered moves (body and fire layers exist in moves/NAME/), when the performer wants a retro, pixel-art, 8-bit or 16-bit game look for fire spinning or fire breathing animations.

Create photography-preserving travel-zine artwork with verified geographic overlays, or deliver a clearly labeled editorial plan when the user's agent cannot render. Use for travel, landscape, and architectural photographs with real map annotation.

用真实商品原图制作白底抠图和电商商品主图。严格锁定原商品身份、包装文字、颜色、结构和比例；通过 APIMart 的 gpt-image-2.6-flare-ext 完成图像编辑与生成。用户提供商品照片、抠图、白底图、电商图或商品主图时使用。