Skills 目录 · 第 81 页

Optional: run an OWASP security review with gstack /cso when the change touches auth or secrets.

Remove duplicated and unnecessary logic from the changed code with the code-simplifier skill. Run it last in the Verify stage.

Optional: audit requirement-to-test coverage after a phase ships, catching the requirements that no test covers.

Builds and reviews frontend interfaces, responsive layouts, accessibility, and interaction behavior. Activate when the user invokes Vex by name or assigns this persona; the specialty alone does not trigger it.

Audits and fixes SEO, SSR/prerendering, meta tags, Open Graph preview cards, sitemap.xml, robots.txt, PageSpeed/LCP performance, and curated GitHub/MCP/Cursor integrations for vibe-coded websites and SPAs. Use when fixing empty React/Vite root divs and search visibility.

Normalizes messy vibe-coded e-commerce HTML/CSS/JS storefronts into clean, light-themed, responsive Zero-CLS production sites with verified catalog imagery and passing DOM tests. Use when cleaning up multi-iteration storefront code and fixing layout shifts.

Create on-screen text hooks and captions for short-form video clips. Built around the Complementarity Principle - on-screen text should ADD to what the viewer hears, not repeat it. Includes podcast clip workflow, hook categories, and Triple Word Score optimization.

适用于视频策略、脚本、YouTube SEO、短视频流水线（Reels、TikTok、Shorts），或在编写脚本前将长内容转化为视频。按平台设定钩子、结构和内容再利用方式。

从YouTube及其他平台下载视频，用于离线观看、编辑或归档。支持多种格式和质量选项。

AI-assisted video editing workflows for cutting, structuring, and augmenting real footage. Covers the full pipeline from raw capture through FFmpeg, Remotion, ElevenLabs, fal.ai, and final polish in Descript or CapCut. Use when the user wants to edit video, cut footage, create vlogs, or build video content.

実写素材のカット、構築、強化のためのAI支援ビデオ編集ワークフロー。生の撮影素材からFFmpeg、Remotion、ElevenLabs、fal.aiを経て、DescriptまたはCapCutで最終仕上げを行う完全なパイプラインをカバーする。ユーザーがビデオの編集、素材のカット、vlogの作成、またはビデオコンテンツの構築を望む場合に使用する。

AI-assisted video editing workflows for cutting, structuring, and augmenting real footage. Covers the full pipeline from raw capture through FFmpeg, Remotion, ElevenLabs, fal.ai, and final polish in Descript or CapCut. Use when the user wants to edit video, cut footage, create vlogs, or build video content.

AI辅助的视频编辑工作流程，用于剪辑、构建和增强实拍素材。涵盖从原始拍摄到FFmpeg、Remotion、ElevenLabs、fal.ai，再到Descript或CapCut最终润色的完整流程。适用于用户想要编辑视频、剪辑素材、制作vlog或构建视频内容的情况。

Composite a graphic onto video — text, a logo/image, a static transparent PNG, or an animated effect — and place it accurately, including locking it to a moving subject (a face, eye, hand, object). Use whenever the user wants to add/overlay/burn-in/stick something onto footage, place a watermark or label, or position a graphic precisely. Covers transparency (RGBA), animated overlays, timed windows, and coordinate-finding.

Create cinematic product videos from frontend projects or webpages with Remotion, real page captures, shot cards, the Ink Press template, beat sync and sound design. Use for full product promos, reproducing Ink Press, single shot-card effects, or finding, adding, replacing and aligning sound effects in an existing product video or shot, including candidate-only SFX searches. 用真实页面、镜头卡与 Ink Press 模板制作产品宣传片；用户要求用 video-shotcraft 做视频、复刻模板、制作单镜头，或给已完成的产品片/镜头找音效候选、加音效、换音效、对齐声音时使用。

Join two clips with a visible transition effect (wipe, slide, dissolve, pixelize, circle open/close, fade-to-black, etc.) using ffmpeg xfade. Use whenever the user wants a transition between shots/scenes, a scene change with an effect, or anything beyond a hard cut. For a barely-visible seam between adjacent cuts use the splice command instead; for generative effects on the footage itself use grok-video-edit.

视频与音频的查看、理解与行动。查看：从本地文件、URL、RTSP/直播源或实时录制桌面获取内容；返回实时上下文和可播放流链接。理解：提取帧，构建视觉/语义/时间索引，并通过时间戳和自动剪辑搜索片段。行动：转码和标准化（编解码器、帧率、分辨率、宽高比），执行时间线编辑（字幕、文本/图像叠加、品牌化、音频叠加、配音、翻译），生成媒体资源（图像、音频、视频），并为直播流或桌面捕获的事件创建实时警报。

ビデオとオーディオの表示、理解、アクション。表示：ローカルファイル、URL、RTSP/ライブストリーム、またはリアルタイムのデスクトップ録画からコンテンツを取得し、リアルタイムコンテキストと再生可能なストリームリンクを返す。理解：フレームを抽出し、ビジュアル/セマンティック/時間的インデックスを構築し、タイムスタンプと自動クリップでモーメントを検索する。アクション：トランスコードと正規化（コーデック、フレームレート、解像度、アスペクト比）、タイムライン編集（字幕、テキスト/画像オーバーレイ、ブランディング、オーディオオーバーレイ、吹き替え、翻訳）、メディアアセットの生成（画像、オーディオ、ビデオ）、ライブストリームまたはデスクトップキャプチャされたイベントのリアルタイムアラートを実行する。

使用 VideoDB Python SDK 摄取、索引、搜索、编辑和监控视频与音频——可从文件、URL 或 RTSP 流上传，构建语音索引和场景索引并支持带时间戳的搜索与可播放片段，进行转码和重新构图，执行时间线编辑（字幕、叠加、配音），并对直播流或桌面捕获运行实时告警。适用于视频搜索、转录、剪辑、转码、流媒体或实时视频告警相关工作时。

Viết và sửa văn tiếng Việt cho thuận mồm, dễ hiểu, không có mùi văn máy hay giọng văn mẫu. Dùng mỗi khi viết hay sửa văn bản tiếng Việt dài hơn vài câu, kể cả khi người dùng không nhắc tới giọng văn, ví dụ bài giảng, tài liệu, README, báo cáo, báo cáo lỗi, văn bản hành chính theo Nghị định 30/2020, hồ sơ dự thầu (thuyết minh đề xuất kỹ thuật E-HSDT), email, lời giải thích trong chat, chữ trên sơ đồ, tệp Word. Dùng ngay khi người dùng chê văn "đọc là biết AI viết", "máy móc", "rời rạc", "văn mẫu", hoặc nhờ humanize, viết theo giọng một người. Kèm script cài skill, bộ nhớ dùng chung cho Claude Code và Codex, học từ phiên cũ, giữ mật khẩu cho agent.

Vinted API: search second-hand listings on any of 27 Vinted country sites (France, UK, US, Germany, Poland, Spain, Italy, the Netherlands, Australia and more) by keyword, category, brand, condition, size, color, material and price range, 96 items per page with price, the Vinted fee and the total, read one listing in full with its category path, photos and sold or reserved state, pull a seller's public profile, listed items and reviews, and look up brand ids, filter options with live counts, the localized category tree and search autocomplete. 9 endpoints, 1 credit each. The official Vinted API serves only allowlisted Vinted Pro businesses and has no marketplace search.

Baixa os tabloides do Supermercados Violeta da praça São Paulo (SP), lendo a galeria de cada post do tabloide. Use quando o usuário mencionar "baixar encartes Violeta", "encartes Violeta" ou colar a URL supermercadovioleta.com.br/tabloide/.

擅长构建可分享、易传播的生成器工具——名称生成器、测验制作器、头像创建器、性格测试和计算器工具。涵盖分享心理学、传播机制，以及构建人们忍不住分享给朋友的工具。适用场景：生成器工具、测验制作器、名称生成器、头像创建器、传播工具。

通过 OCR 将签证文件图像（银行存款、就业、收入和退休证明；HEIC、PNG 或 JPG）翻译成英文，并生成双语 PDF，将原始图像与格式化的认证风格译文配对。适用于签证申请需要将文件翻译成英文、官方证书需要 OCR 并翻译，或移民材料需要双语翻译 PDF 的情况。

Use when building Vision Pro apps — spatial computing, RealityKit 3D content, immersive spaces, or volumes.

让网站图像呈现艺术指导感而非平庸效果——选定一种视觉风格，获取或生成图像（照片、AI 图像、3D 渲染、插画），用一致的裁剪与色彩处理统一它们，并通过蒙版、图层、图像效果和带纹理的背景来呈现；导出经过优化的响应式图像。适用于用户网站已有图像、照片、插画、图标、背景、主视觉或图库，或用户表示图像显得平淡/像图库素材/"gambarnya biasa"，需要为某个区块配图，或想要 AI 图像提示词时。

The BEAUTIFUL family of the visual-reasoning stack: data visuals that maximize communication plus beauty, the sibling pole to the CLEAR family's communication plus clarity. Use when the user asks for something beautiful, striking, data art, a poster, a wall or lobby display, an exhibition piece, a screensaver, a hero image, a glowing network, a LiDAR city, a wind or flow animation, or "make it gorgeous"; and when the user wants options, to make both poles plus a combined candidate and rate them (the default when a request names neither pole). Keeps the truth floor (real data, disclosed mapping) while relaxing style rules.

Arrange encoded marks into a legible artifact with an intentional visual hierarchy, clean layout, and complete annotation (titles, labels, captions, legends, sourcing, alt text). Use after visual-encoding to compose the final figure/slide/panel layout. For v0.1 this skill also owns annotation.

Orchestrator and judge for the visual-reasoning stack. Use to run an end-to-end visualization task (routing through intent → encoding → composition → scientific-visualization → render-qa) OR to critique/grade an existing visual against the failure taxonomy and rubric. Invokes the other skills; it does not duplicate their rules.

Turn an intent record into concrete visual encodings: which data field maps to which channel (position, length, angle, area, color), the exact chart mechanics, and the color strategy. Use after visual-intent and before visual-composition, whenever channel or chart-type choices must be made honestly and legibly.

The front door of the visual-reasoning stack. Use FIRST, before any chart is drawn, whenever the task is to communicate data visually (a chart, figure, slide graphic, map, diagram, or infographic). Establishes communication intent, audience, the abstract data + task, and selects a chart family from the analytic relationship — so downstream encoding/composition choices are derivable, not guessed. By default hands off to both the CLEAR and the BEAUTIFUL family (visual-beautiful).

Divergent orchestrator: given one problem, produce a small portfolio of genuinely-different visualizations — each answering a DIFFERENT analytical question about the same data, each individually valid, each shipped with a trade-off card. Use when the ask is "show me the alternatives / what are the ways to look at this / a few views of X", or whenever a single figure would hide legitimate competing readings. Runs generate → validate → select → characterize. It sits ON TOP OF visual-critic (which validates and scores each member); it does not replace or duplicate the single-figure machinery.

Front door for slide decks and animated HTML pages made to present or communicate: slides in PowerPoint, Keynote, Google Slides, markdown outlines or HTML, builds and transitions, Morph or Magic Move, animated charts for a talk, scrollytelling and stepped explainer pages. Use whenever the user says "make slides", "draft the deck in markdown", "fix my titles", calls a deck cryptic or too mathy, asks whether a transition or build should be animated, wants a chart animated for a talk, or is building a scroll-driven story page. Decides first whether motion earns its place (default: static), then applies the claim-title slide rules and the motion.* principles.

真正查看正在构建的网站——移动端和桌面端截图、开场动画和滚动的胶片条——然后像艺术总监一样依据“基础 vs 高级”评分标准进行批评，并反复迭代，直到它不再显得千篇一律。适用于构建或更改任何页面、区块、主视觉、图像处理或动画之后；每当用户说它看起来基础、平淡、无聊、“kurang wow”、“masih biasa”、“kayak template”、“belum keren”，请求对设计提供反馈，或在宣布任何视觉工作完成之前。

Renderer-independent quality gate. Inspects the ACTUAL rendered artifact at final size in its real medium, applies the hard-constraint gates and accessibility checks, detects visual defects (collisions, clipping, contrast, color-only cues), and prescribes repairs. Use as the last step before any visual is shipped, regardless of what tool produced it.

Vite构建工具模式，包括配置、插件、HMR、环境变量、代理设置、SSR、库模式、依赖预打包及构建优化。在处理vite.config.ts、Vite插件或基于Vite的项目时激活。

Vite构建工具模式，包括配置、插件、HMR、环境变量、代理设置、SSR、库模式、依赖预打包及构建优化。在处理vite.config.ts、Vite插件或基于Vite的项目时激活。

用于TypeScript项目（Node、bun、React/Next.js、Effect）中的Vitest：编写、运行或调试单元/组件测试、模拟、测试工具及覆盖率。

语音智能体代表 AI 交互的前沿——人类与 AI 系统自然对话。挑战不仅在于语音识别与合成，还在于在低于 800ms 延迟下实现自然对话流，同时处理打断、背景噪声和情感细微差别。本技能涵盖两种架构：语音到语音（OpenAI Realtime API，延迟最低、最自然）和流水线（STT→LLM→TTS，控制更强、更易调试）。关键洞察：延迟是约束条件。Hu

Expert in building voice AI applications - from real-time voice agents to voice-enabled apps. Covers OpenAI Realtime API, Vapi for voice agents, Deepgram for transcription, ElevenLabs for synthesis, LiveKit for real-time infrastructure, and WebRTC fundamentals. Knows how to build low-latency, production-ready voice experiences. Use when: voice ai, voice agent, speech to text, text to speech, realtime voice.

Write in the Pirate Wires style - authentic, conversational, contrarian. This voice is direct and confident with irreverent humor, takes obvious-but-unsaid positions, and sounds like a smart friend explaining something they've figured out. Use with anti-ai-writing for tech commentary, cultural criticism, or any content requiring a strong point of view.

适用于交付变慢、工程招聘失效、团队结构不清，或需要决定何时增加工程经理的情况。运行 delivery_throughput_analyzer.py（DORA）、eng_hiring_funnel_calculator.py 和 eng_team_structure_designer.py。

就交付速度、工程招聘、团队结构或值班安排对计划提出质疑。在被要求时或使用 /c-level-agents:vpe-review 时使用。

VR/AR 开发原则：舒适度、交互和性能要求。

Build reactive user interfaces with Vue 3: Composition API, reactivity, components, state, and tooling. Use for any Vue-based UI work.

Vue.js 3 Composition API 模式、组件架构、响应式最佳实践、Pinia 状态管理、Vue Router 导航以及 Nuxt SSR 模式。适用于 Vue、Nuxt、Vite 或 Pinia 项目。在构建或审查 Vue 3、Nuxt 或 Pinia 代码时使用——涉及 Composition API、响应式或路由导航。

高级漏洞分析原则。OWASP 2025、供应链安全、攻击面映射、风险优先级排序。

Sweep all recent Claude Code sessions for ones waiting on the user. Use when the user says "sweep my sessions", "what is waiting on me", "which sessions need me", "scan my sessions", or asks to collect open questions across sessions and answer them. Finds sessions that ended on a question, a decision, an approval or a stuck permission prompt, asks the user, drafts each reply for confirmation, then sends it to that session (desktop app) or hands back a resume command (CLI).

Walgreens product data API: search walgreens.com retail products with regular and sale price, unit price, coupons, rewards and rebates, read one product with UPC, ingredients and offers, find Walgreens and Duane Reade stores near a ZIP, and get live unit counts of one product at up to 20 stores in one call. 4 endpoints; search, product and stores 2 credits, stock 4.

Show work as an interactive picture (flow, cloud or Kubernetes map, sequence, state, timeline…) that the user clicks to ask, correct or confirm. Three modes - plan a feature as small steps and build it step by step, explain how existing code works ("explain the login flow in project A"), or compare designs for something new and pick one. Use when the user wants a visual plan, explanation or comparison, wants to follow along before you build, mentions walkboard, or runs /walkboard.