Image

Close-Up makes or refreshes the page police ask for when a relative with dementia goes missing, with a recent close-up photo the caregiver picks from the NAS photos their account can open, a description, likely places and what changed since the last page. Use when a family caregiver asks to make or refresh that page, says the person's look or condition has changed, or says the person is missing now.

专业服装美术/时尚造型/真人写真提示词写作助手。用于为AI绘画与生图模型(Midjourney、SD、Flux、即梦、可灵、Nano Banana、GPT-image等)撰写高质量的成年人物服装、配饰、首饰、面料、廓形、姿势、场景、光线、镜头提示词,也可用于小说/漫画/游戏人设的服装描写。学习巴黎世家、华伦天奴、圣罗兰、Prada、Miu Miu、The Row、麦昆、Versace、Dior、Chanel、Bottega Veneta、Loewe、Mugler、Schiaparelli、Alaïa、Jacquemus等品牌语言,覆盖针织开衫、毛衣裙、旗袍/新中式、丝袜与鞋履、蕾丝、真丝睡裙、泳装、职业装、礼服露背等品类。支持:手机随手拍/CCD直闪/胶片等真实感写法、姿势与人体结构锁定、批量随机池、角色一致性锁定、COS写真。只要用户提到写真、人像、男装、男模、西装、男性造型、服装描写、穿搭、造型、人设、lookbook、生图提示词、"帮我写一套衣服/一张图"、"有魅力/性感/高级感/氛围感的穿着",即使没有明确说"skill"或"提示词",也要使用本skill。

Turn phone photos or rough scans of book pages and academic papers into a clean, flat, print-like PDF for reading. Straightens curved lines and tilted pages by the text itself, removes fingers, spine shadows, page curl, neighbouring-page slivers, smudges and margin pen notes, keeps printed text, figures, captions and page numbers, and evens out the margins. Use when someone has a photographed or badly scanned book chapter, article or paper (PDF or images) that is hard to read, skewed, curved, stained or off-centre.

Make sure every picture on a website has a proper credit. Use when the user wants to credit photographers, find which pictures have no credit or attribution, work out who made a picture or what license it is under, fix or complete image attributions, build a photo credits page, add "Photo by" captions, or check image credits before a launch.

Build scientific/technical diagrams and concept icons with POV-Ray - atoms, bonds, crystal/molecular structures, vector-field arrows, oscillating "wave" connectors, glowing/energetic effects, metallic finishes, and clean transparent-background renders for dropping into papers, presentations, posters and grant figures. Use this whenever the user wants a diagram of a molecule, crystal lattice, particle/atom arrangement, or a small conceptual icon representing a physical process (light/laser pulses, current/particle flow, heat/thermal disorder, an abstract force or interaction between objects, a vector field) - even if they don't say "POV-Ray" or "3D render" explicitly, e.g. "can you make a little icon for X", "I need a schematic showing Y", "render this crystal structure", "something to represent Z happening between these atoms". Also use it to review or extend an existing POV-Ray scene, or to debug a POV-Ray render that isn't behaving (blank/opaque background, clipped geometry...

Produce publication-quality 2D scientific figures with gnuplot from simulation or column data. Use whenever the user wants to make, plot, or render a figure/graph for a paper, talk, thesis or poster — especially magnetism plots like hysteresis loops (M vs H), temperature-dependent magnetisation (M vs T), and time series (M vs t), including choosing a colour scheme, applying journal/presentation styling, adding axis labels and units, and fitting curves. Trigger on: "make a figure", "plot", "graph", "hysteresis loop", "M vs H / M vs T", "magnetisation curve", "gnuplot", "publication figure", "with viridis/magma/plasma/cividis/turbo/skyrmion/bluegold colours".

Gera e estrutura prompts profissionais de fotografia comercial de moda para e-commerce em formato de colagem (manequim invisível + close-ups macro de textura e barra) a partir de imagens de referência.

Make a rigged Live2D model (a .moc3 bundle with physics, expressions and motions) from one full-body anime drawing, with no Cubism Editor and no hand rigging, using this kit's scripts - candidate drawings with Qwen Image, cut-out, upscale, See-through layer split, layer fixes, rig and build, then checking in a viewer. Use this whenever the user wants a Live2D model, avatar, VTuber model or rig of a character, wants another character put through this pipeline, says something looks wrong with a model made here (a second jaw line, a gap at the waist, eyes or irises, arms, knees, hair, a tilted head), or wants its expressions, motions or the way it moves changed - even if they only say "make her a model" or "fix her eyes".

Convert uploaded images, photos, statues, animals, or design concepts into Rikyu-style geometric crest and Japanese kamon construction blueprint posters in pure SVG. Use when the user uploads an image to turn into an SVG geometric sketch, or asks to replicate or create a Rikyu-style emblem poster, compass-and-straightedge logo blueprint, or golden-ratio geometric grid illustration with construction circles and extended tangent rays.

Generate an image from a text description using Cloudflare Workers AI (the flux-1-schnell model). Use this skill whenever the user wants to create, generate, make, or draw an image / picture / logo / icon / thumbnail / wallpaper / banner / cover from a text idea — even if they don't mention Cloudflare by name. Claude expands the user's request into a strong image prompt, runs the bundled generate.py, and saves a JPG to disk. Trigger on phrases like "genera una imagen de", "crea una imagen", "hazme un logo", "necesito un thumbnail", "generate an image of", "make me a picture of", "create a thumbnail for", "cloudflare image", or a plain description paired with a request for an image. Requires CF_ACCOUNT_ID and CF_API_TOKEN in the project's .env — if missing, this skill tells the user exactly how to set them up before generating.

Cria sprites, tilesets e animações de pixel art grandes e detalhadas no Aseprite pelo servidor pixel-mcp. Use quando o usuário pedir sprite, personagem, tile, ícone, cena de pixel art, spritesheet, ciclo de animação ou um estilo específico (8-bit, 16-bit, GBA, HD pixel, dark fantasy, cyberpunk, isométrico). Also triggers on English requests for sprites, pixel art, sprite sheets and animation cycles. Impõe níveis de resolução, paleta primeiro, direção travada, camadas e revisão visual para que o resultado não fique pequeno e chapado.

Design campaign posters (Instagram, stories, print) the way a senior graphic designer would. Build each poster in HTML/CSS/SVG with real layering, gradients, shapes and light, use photos only as ingredients (stock, ChatGPT or Higgsfield), check every poster at full and phone size, and export PNGs. Use when someone asks for posters, flyers, social graphics or a campaign set for a brand.

Generate extremely simple, cute, personified square character images with rounded heavy forms, two purposeful character colors, one solid background color, and a dominant lower-corner composition. Use when creating an animal, creature, robot, ghost, plant, object, or other character image, including when the agent should infer three product-relevant directions and propose six independent candidates for approval.

Recreate any photo (a selfie, portrait, product shot) as tens of thousands of soft colour blobs trained from random noise by gradient descent, and output the painting, a photo-vs-code comparison, a zoom proof and a noise-to-picture GIF. Use when someone asks to rebuild, "recreate with code", or "gradient descent" an image, or wants a "which one is the photo?" reel.

Turns any grey 3D scene (a .blend with several, one or no cameras, a glTF/GLB, OBJ, FBX, USD, PLY or STL file, a JSON scene spec of primitives, or an idea written as a spec first) into the same scene textured from several calibrated views of one generated photoreal multi-view sheet, so it holds up when the camera moves: exact clay views and passes (object ids, camera-axis depth, world position, normals) from N cameras, one multi-reference image edit of the clay grid with reference images and every full-resolution clay view, per-view refinement accepted by a fixed rule, a measured cross-view consistency score, visibility-aware fusion of all views into each object's own UV atlas scored by leave-one-view-out and seam visibility, optional render-then-refine and silhouette-carving refit rounds, and two outputs: a baked-appearance diagnostic .blend that says it contains lighting and an editable, relightable PBR .blend whose base colour is the projected photo times a smooth correction...

Client-brief coverage and complementary photo shortlists with evidence, unresolved shots, and a reviewed delivery report, for when a photographer asks which selected images satisfy a brief and what is still missing.

Multimodal generative toolkit using Google Gemini native models for image generation and conversational editing (Nano Banana), video generation and editing (Gemini Omni Flash), music creation (Google Lyria 3), and speech synthesis (Gemini Text-to-Speech).

Create coherent multi-image marketing poster series from one selling point. Each poster is an independent image using the Second World visual system: preserve the real source photograph above and naturally extend an element from it into a symbolic second world below.

Lee y construye imágenes con método, el cómo (técnica) y el porqué (historia), el encuadre como base, composición en tres capas de profundidad (primer término, término medio, fondo) frente al plano de sujeto, y la luz al servicio de la composición, y lo traduce a un prompt en inglés para video o imagen con IA, a una escena de Blender o a notas de dirección. Úsala SIEMPRE que el usuario quiera entender por qué una imagen "se ve de cine" o "se ve plana", analizar un plano o una película, componer un frame, un anuncio o una escena de Blender con profundidad, decidir cuántos close-ups usar, o cuando aparezcan palabras como alfabetización visual, visual literacy, lenguaje visual, profundidad, capas, foreground, background, "se ve plano", "se ve barato" o "más cinematográfico", aunque no lo diga así.

Logo, app icon and store-art process - brief, image-model rounds, a versus sheet that tests finalists at real sizes, then one drawing in code rendered to every launcher, adaptive, notification, splash, favicon, font and store asset. Use when choosing a logo, producing app icons, or making store screenshots and posters.

PNG to SVG — ubah gambar PNG (logo, ikon, ilustrasi datar, line art, bentuk bergradasi) menjadi SVG vektor bersih, satu file atau satu folder, dengan pilihan varian warna (original / white / black) plus favicon dan ikon aplikasi. Pakai saat diminta convert, trace, atau vektorisasi PNG ke SVG.

Make a presentation in a calm keynote style as a PowerPoint (.pptx) file - one idea per slide on a dark (or light) page, big numbers in one accent colour, sourced claims, photos graded alike, and the whole script with timings in the presenter notes. Use when the user wants a pitch deck, demo day deck, talk, lecture or class slides "like a keynote" or "like Apple's", or asks for slides like the Ijwi deck. Writes a JSON spec, builds it with python-pptx, checks it, and draws PNG previews of every slide.

Falsifiable design review of an EXISTING UI against the project's own design system. Use when the user says "review this page", "audit the UI", "critique this design", "find defects", "does this look right", or asks for a polish/quality pass on a screen that already ships. Reports findings ranked by user impact with evidence, a concrete fix, and what needs the owner's decision.

通过弹窗引导新手确认屏幕、行业设备及制作范围，制作可选布局色调的工业HMI预览，再展开页面；支持仅调用技能启动、局部编辑及MCGS原生复制，无需MCP或外部供体。