设计与三维建模

Remotion 的最佳实践——在 React 中创建视频。29 条领域专属规则，涵盖 3D、动画、音频、字幕、图表、转场等。在使用 Remotion 在 React 中构建视频时使用——动画、音频、字幕、图表或转场。

Remove the background from any image: upload or link a photo, get a transparent PNG back.

Check generated audio, video and stills for defects. Returns a verdict, not more media.

Design consistent REST APIs: resource modeling, status codes, pagination, versioning, error contracts, and documentation. Use for any API design task.

确保采用 ATS-friendly 的格式，并创建整洁、易于浏览的版式。当用户提到简历格式、版式、模板、设计或修复格式问题时使用。

reverse COMB UUID layout, value discarded

Reviews a PR, diff, design doc or proposal for risks and standards. Use when the user wants something reviewed. Not for release sign-off (ship).

Visual self-review loop for motion graphics — render a draft, build contact sheets at every timeline beat, look at them, critique against a checklist, fix, repeat. Use after any change to a video's visuals or timing, and before calling a video done.

Review a website or web design against the premium-website psychology framework — Halo Effect first impression, cognitive fluency, Peak-End micro-interactions, 2026 trends, and client ownership. Use when designing, auditing, or pre-launch checking a site. Triggers on "/review-site" or "review this site against the premium framework".

Use when writing or reviewing idiomatic Rust — edition 2024 ownership/borrowing design, or fixing LLM pitfalls (clone tax, unwrap infestation, indexed loops). Not for error-type design (rust-error-handling) or SOLID layout (solid-rust).

Explains system design, evaluates tradeoffs, and records architecture decisions and useful documentation. Activate when the user invokes Sage by name or assigns this persona; the specialty alone does not trigger it.

供 skill-tester 作为测试夹具使用的 BASIC-tier 参考技能。统计单词和字符数，并执行基本文本转换。适用于验证 skill-tester 本身，或需要一份最小、已知可用的技能布局来复制时。不是生产技能。

Generate comprehensive edge cases and test scenarios by decomposing a feature or file across 12 risk dimensions, or blast-radius review a change with --diff: what breaks beyond the diff, proven by running real code. Use for pre-implementation risk discovery, QA planning, regression design, exhaustive edge-case enumeration, and change-safety review. Triggers: 'edge cases for X', 'what could break', 'blast radius', 'what could this break beyond the diff', 'test scenarios', 'QA plan', 'risk discovery', 'enumerate failure modes'.

Take a picture of a window, a region or a menu on this machine — through DocShot on macOS, which holds the Screen Recording grant the terminal lacks, and through window-shot.ps1 on Windows — and owns the shared capture and framing scripts every project's capture code calls. TRIGGER when: any screen capture is about to be taken, debugging and measurement included; `screencapture` failed with "could not create image from display"; a capture came back blank or with every window title empty; or a script under this skill's scripts/ is being changed. DO NOT TRIGGER when: the picture is a documentation screenshot to be filed in a project's docs — that is `docs-screenshot`, which runs its own capture scripts and calls these tools directly — or the capture is a browser page the claude-in-chrome tools already screenshot.

Locate macOS screenshots saved to ~/screenshots/ and route them into downstream workflows — most commonly attaching to GitHub PR/issue/comment bodies via the gh image extension, but also useful for previewing, organizing, copying to clipboard, or feeding into other tools. Use when the user mentions "screenshots", "today's screenshots", "find a screenshot", "attach screenshots", "upload screenshot", "images on machine", or wants to locate, preview, or share a local screenshot.

使用 GSAP、ScrollTrigger、SplitText 和 Lenis 平滑滚动构建滚动驱动动画——包括显现、固定区块、横向滚动、视差、擦洗时间线、文字效果——适用于原生 JS、React、Next.js、Vue/Nuxt 或 Astro，并包含正确的初始化、清理和响应式回退。当用户提到 GSAP、ScrollTrigger、Lenis、平滑滚动、视差、固定、擦洗、“animasi saat scroll”、横向滚动或滚动叙事时使用。

构建沉浸式滚动驱动体验的专家——视差叙事、滚动动画、互动叙事和电影级网页体验。类似 NY Times 互动内容、Apple 产品页面和获奖网页体验。让网站感觉像体验，而不仅仅是页面。使用场景：滚动动画、视差、滚动叙事、互动故事、电影级网站。

世界级计算机视觉技能，涵盖图像/视频处理、目标检测、分割和视觉 AI 系统。专长于 PyTorch、OpenCV、YOLO、SAM、扩散模型和视觉 Transformer。包括 3D 视觉、视频分析、实时处理和生产部署。适用于构建视觉 AI 系统、实现目标检测、训练自定义视觉模型或优化推理流水线。

计算机视觉模型：目标检测、分割、训练自定义模型，以及优化或部署推理（YOLO、DETR、SAM、ONNX、TensorRT）。适用于用户想要检测图像或视频中的对象、使用自己标注的图像进行训练，或加速缓慢的视觉模型。

设计实验并分析表格数据：A/B 测试、因果推断、特征工程、模型评估。当用户询问如何设置或解读 A/B 测试、某项改动是否带来了提升、应构建哪些特征，或模型表现如何时使用。

世界级提示词工程技能，涵盖LLM优化、提示词模式、结构化输出与AI产品开发。专长包括Claude、GPT-4、提示词设计模式、少样本学习、思维链及AI评估。包含RAG优化、智能体设计与LLM系统架构。适用于构建AI产品、优化LLM性能、设计智能体系统或实现高级提示技术。

Use when optimizing video SEO — VideoObject schema, YouTube metadata, transcripts, live streams.

为工程技能配置当前仓库：设置问题跟踪器、问题分诊标签词汇和领域文档布局。在首次使用其他工程技能之前运行一次。

Sets up an `## Agent skills` block in AGENTS.md/CLAUDE.md and `docs/agents/` so the engineering skills know this repo's issue tracker (GitHub or local markdown), triage label vocabulary, and domain doc layout. Run before first use of `to-issues`, `to-prd`, `triage`, `diagnose`, `tdd`, `improve-codebase-architecture`, or `zoom-out` — or if those skills appear to be missing context about the issue tracker, triage labels, or domain docs.

管理 shadcn/ui 组件和项目，提供构建现代设计系统所需的背景、文档和使用模式。

Use when defining or auditing shadcn/ui design tokens, OKLCH colors, or dark/light mode CSS variables.

提供使用 shadcn/ui 组件集成和构建应用的专家指导，包括组件发现、安装、自定义和最佳实践。

Create short-form video content (Reels, TikTok, Shorts) using practitioner methodology. Covers production workflow from concept to publish, hook construction, format selection, and algorithm optimization.

Local video studio for your coding agent: scenes, voice, music and captions, on your machine.

分析项目描述或站点需求，并将项目章程综合到 SITE.md 中，供 Stitch Build Loop 使用

Hand-drawn whiteboard for system design mock interviews; your AI watches, asks and gives feedback.

Mine the CURRENT session for how the agent skills were actually used - friction hit, commands corrected, recipes discovered, docs found stale - distill only the SELECTIVE, generalizable improvements, apply them to the relevant SKILL.md / reference files in the skills repo, and ship via vd:ship. Conservative by design: high bar, evidence-backed, no scope creep. Use when the user says 'evolve the skills', 'skill-evolve', 'improve the skills from this session', 'capture what we learned into the skills', 'update the skills based on this session', or runs an end-of-session skill retro. Routes standing-behavior corrections to vd:rule-miner and project/task facts to memory - it only touches skill content.

依据官方规范和最佳实践评估 Agent Skill 设计质量。适用于审查、审计或改进 SKILL.md 文件和技能包时。提供多维度评分和可操作的改进建议。

Frontier-design reference: file formats, converters, definitions, sensors, companies, design jobs.

Transcribe audio and video into speaker-labelled transcripts, subtitles, clips, and cited Q&A.

面向质量优先的软件架构指南。当用户想要编写代码、设计架构、分析代码，或任何与软件开发相关的情况时，应使用此技能。

当用户要求通过 OpenAI 的视频 API 使用随附的 CLI（`scripts/sora.py`）生成、混剪、轮询、列出、下载或删除 Sora 视频时使用，包括“生成 AI 视频”“Sora”“视频混剪”“下载视频/缩略图/精灵图”以及批量视频生成等请求；需要 `OPENAI_API_KEY` 和 Sora API 访问权限。

Procedural sound design with the sfx engine — writing a cue sheet against a video's timeline, which sound goes with which visual event, gain hierarchy, panning, musical key, automatic mastering to -14 LUFS, adding new generators, and verifying sync. Use when adding, changing or checking audio, SFX, music beds or loudness for any video in this repo.

The studio's Sound Designer. Owns the sound of a video for the studio-director, covering dialogue cleanup, music bed and ducking, sound effects synced to picture and motion, the final mix, loudness per platform, and stems, and reports back with a Sound Handoff. Picks and runs audio-edit, sound-design, filler-removal and edl-edit music. Use when the director reaches the sound pass, or when the user asks to fix or mix audio, add music or SFX, make the voice sound better, set loudness for YouTube/Reels/podcast, or says "ظبط الصوت", "الصوت واطي", "حط مزيكا", "مؤثرات صوتية", "ساوند ديزاين".

The studio's Sound Designer. Owns the sound of a video for the studio-director, covering dialogue cleanup, music bed and ducking, sound effects synced to picture and motion, the final mix, loudness per platform, and stems, and reports back with a Sound Handoff. Picks and runs audio-edit, sound-design, filler-removal and edl-edit music. Use when the director reaches the sound pass, or when the user asks to fix or mix audio, add music or SFX, make the voice sound better, set loudness for YouTube/Reels/podcast, or says "ظبط الصوت", "الصوت واطي", "حط مزيكا", "مؤثرات صوتية", "ساوند ديزاين".

Prove you made it first. Blockchain Timestamp creative work via Claude or any MCP-enabled AI Agent.

Investigates a problem or an unclear-impact feature and proposes options. Use when the user has a problem to understand before building. Not for scoped work (design).

当用户要求文本转语音旁白或配音、无障碍朗读、音频提示，或通过OpenAI Audio API进行批量语音生成时使用；使用内置语音运行捆绑的CLI（`scripts/text_to_speech.py`），实时调用需要`OPENAI_API_KEY`。自定义语音创建不在范围内。

适用于用户请求文本转语音旁白或配音、无障碍朗读、音频提示或批量语音生成。OpenAI仍为默认选项；Atlas Cloud是用于异步多语言语音的显式可选后端。自定义语音创建不在范围内。

Spring Boot 架构模式、REST API 设计、分层服务、数据访问、缓存、异步处理和日志。适用于 Java Spring Boot 后端工作。

Spring Boot 架构模式、REST API 设计、分层服务、数据访问、缓存、异步处理和日志。适用于 Java Spring Boot 后端工作。

Spring Boot 架构模式、REST API 设计、分层服务、数据访问、缓存、异步处理和日志。适用于 Java Spring Boot 后端工作。

Spring Boot 架构模式、REST API 设计、分层服务、数据访问、缓存、异步处理和日志记录。用于 Java Spring Boot 后端工作。适用于构建或审查 Spring Boot 后端——REST 层、服务、数据访问、缓存或异步工作。

Spring Boot 架构模式、REST API 设计、分层服务、数据访问、缓存、异步处理和日志记录。用于 Java Spring Boot 后端工作。适用于构建或审查 Spring Boot 后端——REST 层、服务、数据访问、缓存或异步工作。

将此技能用于强化学习任务，包括训练 RL 智能体（PPO、SAC、DQN、TD3、DDPG、A2C 等）、创建自定义 Gym 环境、实现用于监控和控制的回调、使用向量化环境进行并行训练，以及与深度 RL 工作流集成。当用户请求 RL 算法实现、智能体训练、环境设计或 RL 实验时应使用此技能。