Image

在 Wisp（Astro 技术博客主题）站点中写作的 skill。用户要求写、改写、扩写一篇博客文章，把笔记、论文、文档或对话整理成博客，批量生成文章草稿，或者检查文章的 frontmatter、格式、标签是否合规时使用。提供文章模板、创建脚本、检查脚本和中文技术写作规范，保证文章风格与主题一致。画图请交给 wisp-diagram skill，本 skill 不负责。

Generate images with the user's own API keys (Google Gemini "Nano Banana" or OpenAI gpt-image). Use when asked to generate, create or make an image, illustration, icon, cut-out, sticker, thumbnail, texture, mockup or visual asset, to edit an image, or to make several images that stay consistent (same character or object). Also used by the vox-explainer skill for collage cut-outs.

Use when the user wants an architecture overview, system structure, "how is this project organized", or C4-style diagrams of a codebase. Reads the code and writes docs/codemap/architecture.md with three zoom levels as Mermaid diagrams.

Use when the user wants a project structure tree, a "map" of features and files, or the big picture of a project at a glance (including when they say mindmap). Writes docs/codemap/mindmap.md as a nested tree list of features, files and key functions, not as a diagram.

服务架构图双向转换。方法一 nl2mmd：自然语言描述的服务器/服务/子服务/调度关系 → 约定方言的 Mermaid；方法二 mmd2html：该方言 Mermaid → 终端风交互 HTML（可拖拽、收合上浮、Alt 全显把手、localStorage 记忆布局）

Image jobs through the Codex CLI (gpt-image-2, ChatGPT login) that get stuck or go wrong - codex-imagegen hanging, "Reading additional input from stdin", the wrong image under a filename, identical outputs, a reference ignored or a new room invented, quota burnt on retries; batches and chained edits. One safe runner script. One-off icons without a project can stay with the codex-bridge plugin's generate-image and edit-image skills.

Acrescenta a uma aula UM visual correto e mínimo — diagrama (Mermaid), figura ou gráfico (SVG) ou imagem de referência — seguindo os princípios da aprendizagem multimídia, e confere o resultado olhando a página renderizada. Use durante uma aula quando uma ideia fica mais clara vista do que lida — estrutura, processo, causa e efeito, comparação, linha do tempo, espaço, dados ou uma obra de arquitetura/design.

Teach a subsystem, change, or idea so the person actually understands it: runs how for the mechanics and its History view for the reasons, then explains in plain layers with diagrams that build up.

Turn an article, talk, thread, paper or notes into a one-page hand-drawn infographic in Excalidraw's style - a title, six or so panels that each make one point with a small diagram and a one- or two-sentence takeaway, and a source line - delivered as a PNG to post and an .excalidraw file to edit by hand. Use when the user wants a sketch page, a visual summary, an Excalidraw-style explainer or cheat sheet, "one of those hand-drawn infographic posts", or a page like an example image they share.

The Graphic Designer role: hands-on craft of visual communication. Grids and layout (column and modular grids, A4 setup, image-text composition), typography fundamentals (size ratios, weight contrast, alignment, rivers, widows), experimental typography (outline surgery, blending, constraints, physical experiments), poster languages (brutalism, grunge, texture, threshold, displacement), logo construction (types, shape psychology, vectorising, stress tests, delivery) and file handover hygiene. Use when executing or critiquing the craft of a poster, layout, cover, typographic piece or logo, picking type or a grid, or explaining why a layout looks amateur. Works inside a direction set by the art director and a concept set by the creative director; hands UX/UI screens to ux-ui-expert.

Publish or update pages, reports, diagrams and assets in a GitHub repository Wiki, preserving existing content and verifying the published result. Use for GitHub Wiki publishing or sync requests, rather than GitHub Pages deployment or general documentation generation.

Turn a picture of a character the user likes into an animated Kobutai stage character - analyse the upload, draw a multi-view profile sheet, generate idle/performance/reaction GIFs with GPT Image, and package a spritesheet + pet.json the Kobutai extension (and Codex pets) can import. Use when the user wants to make, hatch or import a Kobutai character, pet or GIF from an image or idea. Not for editing Kobutai's code.

Independently critique, audit, and test an existing website, app, screen, screenshot, HTML mockup, prototype, or frontend implementation. Use for UI or UX reviews, usability issues, interaction bugs, visual hierarchy and layout audits, accessibility and responsive checks, prioritized recommendations, and comparison of design revisions. Distinguish visible evidence, browser-verified behavior, inferred risks, and unknowns. Do not use as the primary UI-building workflow; use frontend-workflow to build or change an app's UI.

Render an interactive diagram or chart from a few lines of text, inline in T3 Code or as an HTML page: flow, DAG, sequence, state machine, call tree or call diff; timing of a race, gantt of where latency goes, latency histogram, retry backoff, sankey of a traffic split, fishbone of incident causes, wait-for graph of a deadlock, kanban of work items, stock-and-flow of a queue over time. Use for viz, visualize, graph, chart, or any diagram.

Build or redesign a whole WordPress site from a visual reference — brief, local HTML mockup, conversion to an editable block theme, local staging with replayable tools, a client-approved deployment and five measured acceptance gates — for showcase sites, WooCommerce stores (block cart and checkout, filters, consent, payments, newsletter, legal pages) and web apps. Use this whenever the user wants to create, rebuild or replace a WordPress site or theme, leave a page builder such as Elementor or Divi, or sends a site brief with a reference image or URL.

Оцифровывай фотографии тетрадей, книг и бумажных документов в аккуратный PDF: разделяй развороты, исправляй перспективу, плотно подгоняй границы, сохраняя записи и чертежи. Применяй также для исправления полей такого PDF; не для перепечатки текста, OCR как основной задачи или генеративной реставрации.

为博客、技术文档绘制 Wisp 风格的概念图、架构图、模型结构图、算法示意图（自注意力、Transformer、矩阵乘、KV cache、MoE、RAG、训练循环等）。当用户或 Agent 需要画图、生成 .excalidraw / .drawio 源文件、导出可移植 SVG/PNG，或提到 excalidraw、drawio、draw.io、Wisp 风格图表、语义色（Q/K/V）、图表调色板时使用。不适用于数据图表（折线图、柱状图等）。

Use when asked to plan a feature or project before building it, design its architecture, or slice a large piece of work. Produces a local markdown plan centered on architecture diagrams and interface definitions, sliced for large work and kept to one page for small work, stress-tested by a critic/defender debate, and presented as an Artifact for review. Implement it afterwards with `implement-plan`.

Work a design out together with the owner when they say "let's work through the design", "design session", "explore options for...", "I want to think through..." or "let's work it out together". The owner leads; the agent offers 2-3 alternative shapes with diagrams, traces consequences (including the additive check against later milestones), catches contradictions and offers a throwaway spike when only running something will settle a question. Notes go to a scratch file, not the system design. Ends with a promote step: the owner picks what enters the system design as settled or working, what is parked for a later milestone and what is dropped; the edit is committed and the notes file deleted. Use any time: before a system design exists, during milestone planning or mid-milestone. Not for drafting a whole system design (use system-design) or planning a rung (use plan-milestone).

Read and change Figma designs from the shell, without an MCP server, through the `figma-access` CLI. From a Figma URL and a personal access token it reads any file through the REST API (frames, exact node styles, rendered images, original image fills), with no Figma Desktop needed; with the Desktop Bridge plugin running in Figma Desktop it also runs Figma Plugin API code in the open file, exports screenshots, reads the selection, console logs and changes, and manages variables, components, FigJam boards and Slides. Use when the user shares a Figma link or asks to inspect, implement in code, export, create, edit, audit or screenshot something in Figma, to work with Figma variables, tokens or components, or names this skill.

Draws diagrams, charts, and technical illustrations (44 types) as self-contained HTML with inline SVG, plus optional SVG/PNG/PDF export, in the project's tracked color and font theme. Use this skill whenever the user asks to draw, diagram, visualize, chart, map, sketch, or illustrate something, even without saying 'illustrate': architecture or system design, infrastructure, cloud or deployment topology, data flow, flowcharts, sequence, state machines, ER or database schema, UML class, dependency graphs, org charts, timelines, Gantt, swimlanes, process or workflow maps, user journeys, kanban, story maps, quadrants, Venn, pyramids, fishbone root-cause, Sankey, Wardley maps, bar, line, scatter, heatmap, waterfall, treemap, radar, or polar charts, and exploded or axonometric views. Also use to redraw .drawio, Mermaid (.mmd), or .excalidraw files, add a diagram to docs, a PR, or a spec, set up the illustration theme, or export a diagram. Not for photos or AI-generated raster images.

Production-grade prompts for any major image generation/editing model: Nano Banana (Gemini Image), GPT Image 2.5 (Flare/Sunburst) and earlier, DALL·E 3, Seedream, Midjourney v6/v7 + Niji, SDXL/SD3.5, FLUX, Imagen, Ideogram, Recraft. Use for generating or editing images, illustrating articles/videos/podcasts, remixing templates, identity preservation, multi-image fusion, non-destructive edits, isometric/bento layouts, in-image text, JSON/YAML structured prompts, sprite sheets, or reconstructing a prompt from a reference image. If a model is named, read `references/models/_index.md` then its file. Triggers: image prompt, image generation, nano banana, gpt image, dall-e, midjourney, niji, stable diffusion, sdxl, flux, imagen, ideogram, recraft, seedream, 纳米香蕉.

Create beautiful, self-contained HTML visualizations from any content or idea. Use for: slide decks, presentations, infographics, dashboards, flowcharts, diagrams, timelines, comparison tables, data visualizations, landing pages, one-pagers, org charts, mind maps, process flows, kanban boards, report summaries, or any visual that helps humans digest information faster. Trigger on requests like "visualize this," "make a deck," "create a slide," "build an infographic," "show me a dashboard," "make this visual," or any request to present information in a visual HTML format.

Launch and supervise Orca workers through the canonical Sprite prepare-and-attach entrypoint, with an explicit local route for Opus roots and the rare Fable assignment for the most sensitive work, and reclaim what a settled Dispatch left behind. Use for worker dispatch, native Run messaging, remote diagnosis, ending a done lane, or source return.