docs: add image, compositing, and VFX answers

docs/public/llms.txt

@@ -9,7 +9,7 @@ Languages: [English](https://dcc-mcp.github.io/llms.txt) · [简体中文](https
 - [Homepage and one-prompt setup](https://dcc-mcp.github.io/)
 - [Marketplace search](https://dcc-mcp.github.io/marketplace): Search the official live catalog and copy consent-aware agent prompts or typed CLI installation commands.
 - [Showcase gallery](https://dcc-mcp.github.io/showcase): View real ecosystem outputs and copy the corresponding production-oriented agent prompts.
-- [Common AI + DCC tasks](https://dcc-mcp.github.io/use-cases): Direct answers and copyable prompts for Maya, Blender, random scene generation, and game creation.
+- [Common AI + DCC tasks](https://dcc-mcp.github.io/use-cases): Direct answers and copyable prompts for Maya, Blender, image editing, film editing and compositing, VFX, random scene generation, and game creation.
 - [Agent workflow](https://dcc-mcp.github.io/agents): Install the public Skill, discover typed tools, validate calls, and diagnose failures.
 - [Developer routing](https://dcc-mcp.github.io/developers): Decide whether to build an adapter, a specialized Skill, or a shared Core change.
 - [Ecosystem directory](https://dcc-mcp.github.io/ecosystem): Adapters, extensions, asset providers, AI services, and studio integrations.

docs/public/zh/llms.txt

@@ -9,7 +9,7 @@
 - [官网与一段提示词完成接入](https://dcc-mcp.github.io/zh/)
 - [Marketplace 搜索](https://dcc-mcp.github.io/zh/marketplace)：搜索官方在线目录，复制安全 Agent 提示词或类型化 CLI 安装命令。
 - [案例画廊](https://dcc-mcp.github.io/zh/showcase)：查看真实生态产出，并复制对应的生产级 Agent 提示词。
-- [常见 AI + DCC 任务](https://dcc-mcp.github.io/zh/use-cases)：直接回答如何控制 Maya、Blender、生成随机场景内容和制作游戏，并提供可复制提示词。
+- [常见 AI + DCC 任务](https://dcc-mcp.github.io/zh/use-cases)：直接回答如何控制 Maya、Blender、修图、剪辑合成、制作特效、生成随机场景内容和制作游戏，并提供可复制提示词。
 - [Agent 工作流](https://dcc-mcp.github.io/zh/agents)：安装公开 Skill、发现类型化工具、验证调用并诊断失败。
 - [开发者路由](https://dcc-mcp.github.io/zh/developers)：判断应该构建适配器、专项 Skill 还是共享 Core 能力。
 - [生态目录](https://dcc-mcp.github.io/zh/ecosystem)：适配器、扩展、资产提供方、AI 服务和工作室集成。

docs/use-cases.md

@@ -1,6 +1,6 @@
 ---
-title: AI control for Maya, Blender, and game creation
-description: Direct answers and copyable agent prompts for controlling Maya or Blender with AI, creating random spheres in Maya, and making a game.
+title: AI control for Maya, Blender, image editing, compositing, VFX, and games
+description: Direct answers and copyable prompts for controlling creative software with AI, from Maya and Blender to photo editing, compositing, visual effects, and game creation.
 pageClass: route-page
 ---
 
@@ -56,6 +56,30 @@ Start with the smallest playable loop, then choose the connected engine: [Unreal
 Help me make a small playable game with DCC-MCP. First ask me for the engine (Unreal, Unity, or Godot), target platform, visual style, and one-sentence playable loop. Reduce the idea to one level and one win or fail condition. Inventory the connected DCCs and engine, discover typed tools, and inspect the official Marketplace for any license-safe assets we need. Ask before installing packages, downloading assets, or changing system state. Build the smallest playable slice, run it, validate controls and the win/fail loop, capture evidence, and report the project and packaged-build paths. Do not claim completion without a runnable build.
 ```
 
+## How do I edit photos with AI?
+
+Use the official [Photoshop adapter](https://github.com/dcc-mcp/dcc-mcp-photoshop) for retouching, color correction, masks, layers, and export. The agent should preserve the original, work non-destructively, and ask for the intended look and output format before editing.
+
+```text
+Use the dcc-mcp Skill to connect to Photoshop and help me edit this image. First ask what must change, what must remain untouched, the reference look, and the delivery format. Preserve the original document, use new layers, adjustment layers, masks, or Smart Objects where supported, and never flatten or overwrite the source without my approval. Discover the typed Photoshop tools, make a bounded first pass, compare the result with the brief, validate dimensions, color mode, layer structure, and missing assets, then export to a new path only after I approve the preview.
+```
+
+## How do I edit or composite a film with AI?
+
+Route the job by intent: [Premiere Pro](https://github.com/dcc-mcp/dcc-mcp-premiere) for timeline editing, [After Effects](https://github.com/dcc-mcp/dcc-mcp-aftereffects) for motion graphics and layer-based shot work, or [Nuke](https://github.com/dcc-mcp/dcc-mcp-nuke) for node-based compositing. Do not move a workflow to another host merely because its tool catalog is easier to reach.
+
+```text
+Use the dcc-mcp Skill to help me edit or composite this film. First ask whether the task is timeline editing in Premiere, motion graphics or layer compositing in After Effects, or node compositing in Nuke. Inventory the source media and confirm resolution, frame rate, color space, audio requirements, and delivery target. Preserve source files, create a new project, sequence, or comp, discover typed tools, complete one representative shot or sequence, render a preview, check missing media and frame range, and report the project and preview paths. Ask before installing codecs or plugins, relinking media, or starting a final render.
+```
+
+## How do I create visual effects with AI?
+
+Use [Houdini](https://github.com/dcc-mcp/dcc-mcp-houdini) for procedural geometry and simulation, Nuke or After Effects for shot compositing, and [Unreal Engine](https://github.com/dcc-mcp/dcc-mcp-unreal) for real-time effects. A good agent first identifies where the effect belongs, then builds and validates the smallest representative shot.
+
+```text
+Use the dcc-mcp Skill to help me create a visual effect. First ask for the shot brief, reference, duration, camera, delivery format, and whether the effect belongs in Houdini simulation, Nuke or After Effects compositing, or Unreal real-time VFX. Inventory connected hosts, discover typed tools, and propose the smallest representative setup. Keep controls editable, preserve source plates and scenes, cache or render only when needed, produce a preview, validate timing, scale, color space, and missing dependencies, and report the scene, cache, and preview paths. Ask before downloading assets, installing plugins, or starting an expensive simulation or final render.
+```
+
 ## Why use DCC-MCP instead of asking for a script?
 
 Typed discovery, validated arguments, instance routing, audit evidence, diagnostics, and consent boundaries are reusable across agents. The agent can change; the production contract stays stable. See [For Agents](/agents) for the complete workflow and failure-reporting rules.

docs/zh/use-cases.md

@@ -1,6 +1,6 @@
 ---
-title: 用 AI 控制 Maya、Blender 与制作游戏
-description: 回答如何用 AI 控制 Maya 或 Blender、在 Maya 创建十个随机小球，以及从零开始做游戏，并提供可复制的 Agent 提示词。
+title: 用 AI 控制 Maya、Blender、修图、合成、特效与游戏制作
+description: 回答如何用 AI 控制创意软件，覆盖 Maya、Blender、修图、剪辑合成、视觉特效与游戏制作，并提供可复制的 Agent 提示词。
 pageClass: route-page
 ---
 
@@ -56,6 +56,30 @@ dcc-mcp-cli search --query "检查 Blender 场景" --dcc-type blender
 使用 DCC-MCP 帮我制作一个小型可玩游戏。先询问我选择 Unreal、Unity 还是 Godot，以及目标平台、视觉风格和一句话可玩循环。把想法缩小为一个关卡和一个胜利或失败条件。盘点已连接的 DCC 与引擎，发现类型化工具，并检查官方 Marketplace 是否有需要的安全授权资产。安装软件包、下载资产或改变系统状态前必须先征得我的同意。构建最小可玩切片，实际运行它，验证控制与胜负循环，保存证据并报告项目和打包构建路径。没有可运行构建时不得宣称完成。
 ```
 
+## 我想用 AI 修图，要怎么做？
+
+使用官方 [Photoshop 适配器](https://github.com/dcc-mcp/dcc-mcp-photoshop)完成修饰、调色、蒙版、图层与导出。Agent 应保留原图、采用非破坏性编辑，并在开始前询问目标效果与交付格式。
+
+```text
+使用 dcc-mcp Skill 连接 Photoshop 并帮我修图。先询问需要改变什么、哪些内容必须保持不变、参考效果和交付格式。保留原始文档，在宿主支持时使用新图层、调整图层、蒙版或 Smart Object；未经我批准，不得合并图层或覆盖源文件。发现类型化 Photoshop 工具，完成一次范围有限的初稿，对照需求检查效果，验证尺寸、颜色模式、图层结构和丢失资产；只有我确认预览后，才导出到新的路径。
+```
+
+## 我想用 AI 剪辑或合成片子，要怎么做？
+
+按意图选择宿主：[Premiere Pro](https://github.com/dcc-mcp/dcc-mcp-premiere) 负责时间线剪辑，[After Effects](https://github.com/dcc-mcp/dcc-mcp-aftereffects) 负责动态图形和图层式镜头处理，[Nuke](https://github.com/dcc-mcp/dcc-mcp-nuke) 负责节点式合成。不能因为另一个宿主的工具更容易调用，就擅自迁移工作流。
+
+```text
+使用 dcc-mcp Skill 帮我剪辑或合成这部片子。先询问任务属于 Premiere 时间线剪辑、After Effects 动态图形或图层合成，还是 Nuke 节点合成。盘点源媒体，确认分辨率、帧率、色彩空间、音频要求和交付目标。保留源文件，创建新的项目、Sequence 或 Comp，发现类型化工具，先完成一个代表性镜头或段落，渲染预览，检查丢失媒体和帧范围，并报告项目与预览路径。安装 Codec 或插件、重新链接媒体或开始最终渲染前必须征得我的同意。
+```
+
+## 我想用 AI 做特效，要怎么做？
+
+使用 [Houdini](https://github.com/dcc-mcp/dcc-mcp-houdini)完成程序化几何与模拟，使用 Nuke 或 After Effects 完成镜头合成，使用 [Unreal Engine](https://github.com/dcc-mcp/dcc-mcp-unreal)完成实时特效。Agent 应先判断特效属于制作链的哪一层，再构建并验证最小代表镜头。
+
+```text
+使用 dcc-mcp Skill 帮我制作视觉特效。先询问镜头需求、参考、时长、摄影机、交付格式，以及效果应在 Houdini 模拟、Nuke 或 After Effects 合成，还是 Unreal 实时 VFX 中完成。盘点已连接宿主，发现类型化工具，并提出最小代表性方案。保持参数可编辑，保留源素材和场景，只在需要时缓存或渲染，生成预览，验证时序、比例、色彩空间和缺失依赖，并报告场景、缓存和预览路径。下载资产、安装插件或启动昂贵模拟与最终渲染前必须征得我的同意。
+```
+
 ## 为什么不直接让 AI 临时写脚本？
 
 类型化发现、参数验证、实例路由、审计证据、诊断与授权边界可以跨 Agent 复用。Agent 可以变化，生产契约保持稳定。完整工作流和失败上报规则见 [Agent 使用](/zh/agents)。

scripts/validate-site.mjs

@@ -69,10 +69,10 @@ if (!showcaseSource.includes('navigator.clipboard.writeText')) throw new Error('
 
 const englishUseCases = readFileSync(join(dist, 'use-cases.html'), 'utf8')
 const chineseUseCases = readFileSync(join(dist, 'zh', 'use-cases.html'), 'utf8')
-for (const phrase of ['control Maya with AI', 'control Blender with AI', 'create ten random spheres in Maya', 'want to make a game']) {
+for (const phrase of ['control Maya with AI', 'control Blender with AI', 'create ten random spheres in Maya', 'want to make a game', 'edit photos with AI', 'edit or composite a film with AI', 'create visual effects with AI']) {
   if (!englishUseCases.toLowerCase().includes(phrase.toLowerCase())) throw new Error(`English use cases are missing: ${phrase}`)
 }
-for (const phrase of ['我想用 AI 控制 Maya', '我想用 AI 控制 Blender', '我想在 Maya 创建十个随机的小球', '我想做一个游戏']) {
+for (const phrase of ['我想用 AI 控制 Maya', '我想用 AI 控制 Blender', '我想在 Maya 创建十个随机的小球', '我想做一个游戏', '我想用 AI 修图', '我想用 AI 剪辑或合成片子', '我想用 AI 做特效']) {
   if (!chineseUseCases.includes(phrase)) throw new Error(`Chinese use cases are missing: ${phrase}`)
 }