dsh-visual-skin
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@dsh-visual-skinuse this photo to create a new skin theme and preview it"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
dsh-visual-skin
DeepSeek Harness Skin Studio — A one-click visual skinning tool for DeepSeek Harness
A visual skin studio for DeepSeek Harness (DSH): design, preview and safely apply skins with one click — no config files to touch. Ship as an Agent Skill + MCP server + DSH plugin, with a one-shot PowerShell installer.
After installation, your AI assistant can automatically complete the full skinning workflow of image upload → theme creation → isolated preview → manual confirmation and application with just one image; you can also open the standalone Studio web page to visually adjust the skin and write it to DSH with one click.
✨ Features
Capability | Description |
🎨 Skin Studio (Visual Designer) | Standalone Web UI: dual-zone background (sidebar/main area), image/gradient backgrounds, glassmorphism, token color picker, zone selector, WYSIWYG |
🤖 Agent Skill | After installation, when the DSH / Codex assistant receives a request like "change the skin / use this image as background", it automatically orchestrates the full skinning workflow |
🔌 MCP Server (26 tools) |
|
🖥️ DSH Plugin | Persistent background (body-level CSS, persists across conversations), conversation overlay (takes effect on click, seamless switching), sidebar separator, balance bar (balance button above settings + balance display in conversation details, refreshable) |
🛡️ Safety Boundary | All "write to DSH" operations go through isolated DSH preview → immutable apply plan → manual confirmation in the visible UI; MCP can never forge confirmation |
📦 One-Click Installation | PowerShell installer automatically completes: build → install Skill → register plugin → register MCP → start Studio → (optional) sync Codex |
Related MCP server: dstools
🎬 Demo
Hero · Dark Showcase (12s) | Generate · From Image to Skin (7s) |
|
|
Animation by huashu-design · Storyboard cards at docs/demo/storyboard.md · Direction boards at docs/demo/direction-boards.html
📦 Installation
Prerequisites
Dependency | Version | Description |
Windows | 10 / 11 | Installer is a PowerShell script |
22+ | Required for plugin and Controller to run | |
11+ | Only needed for first installation (to build portable runtime) | |
0.1.0-rc.6 | Target version, capability pinned; plugin fails closed if version mismatch |
Method 1: Command Line Installation (Recommended)
# 1. 获取项目
git clone https://github.com/trrrrrryg/dsh-visual-skin.git
cd dsh-visual-skin
# 2. 一键安装(默认目标:%USERPROFILE%\.dsh,profile: web,Controller 端口 11862)
powershell -NoProfile -ExecutionPolicy Bypass -File .\install.ps1The installer automatically completes:
Build/reuse self-contained portable runtime (Controller + Studio + MCP + plugin, ~51MB, installed under Skill directory);
Install Skill to
%DSH_HOME%\skills\deepseek-harness-skin-studio(DSH assistant immediately gains this skill);Register DSH plugin (
cordis.patch.ymlmanaged block, includesSkin Settingssettings card);Register MCP server (
mcp__skin-studio__*, pointing to the installed self-contained runtime);Start Studio (Controller, default port 11862, browser access
http://127.0.0.1:11862);If Codex is detected, automatically sync Skill and register Codex MCP.
After installation, restart DSH once (or restart in the desktop launcher) to activate the profile patch and MCP client.
Optional Parameters
Parameter | Description |
| DSH root directory (default |
| Target profile (default |
| Studio data directory (default |
| Studio Controller port (default |
| Use already built runtime, skip building |
| Skip corresponding component |
| Automatically restart DSH after installation (default no restart to avoid interrupting running sessions) |
Example (custom installation + auto restart DSH after installation):
.\install.ps1 -DshHome C:\Users\me\.dsh -ProfileName web -ControllerPort 12000 -RestartDshMethod 2: AI Installation
Give this repository to any agent and tell it:
Please read the README and the
install.ps1in this repository, then execute the installation.
The agent will complete the installation following the above process (it can also run install.ps1 directly).
Verify Installation
# 插件健康检查(在 DSH 端口上,应返回 ok:true;默认 DSH 端口 10402)
Invoke-RestMethod http://127.0.0.1:10402/dsh-skin/health
# Controller / Studio 状态检查(应返回 ok:true、capabilities.compatible:true)
Invoke-RestMethod http://127.0.0.1:11862/api/v1/statusStudio opens:
http://127.0.0.1:11862(or click Start menu / tray icon)A "Skin Settings" card should appear in the DSH settings page; say "change the skin" to the assistant to trigger the
mcp__skin-studio__*tools.
🚀 Usage
After installation and restart, simply say to the DSH assistant:
"Give me a new skin" / "Use this image as background" → The assistant automatically calls
mcp__skin-studio__*to create the design and preview it in Studio;"Replace the current skin background" → The assistant modifies the current design;
Manually adjust in Studio (background, zones, tokens, separators) → Click Write to my DSH, manually confirm in the pop-up plan confirmation window → Restart DSH to apply.
Skinning Flowchart
用户指令 / 图片 ──▶ Agent Skill ──▶ MCP (design_create → asset_upload → theme_patch
→ theme_validate → preview_start)
│
▼
隔离 DSH 预览(真实渲染,不碰你的 DSH)
│
▼
Studio 可见确认(theme_apply_plan → theme_apply)
│
▼
你的 DSH 皮肤生效(可随时 Restore 还原)🗂️ Directory Structure
├─ install.ps1 # 一键安装器(命令行 / AI)
├─ apps/
│ ├─ controller/ # Studio 后端(API、隔离预览、apply/restore、GC)
│ └─ studio/ # Studio Web UI(React + Vite)
├─ packages/
│ ├─ dsh-plugin/ # DSH 插件(host 路由 + client 皮肤渲染)
│ ├─ theme-schema/ # ThemeSpec v2 数据契约
│ ├─ design-session-core/ # 设计会话 / 确认票据 / 操作日志
│ ├─ shared/ # 跨端类型与错误码
│ ├─ agent-cli/ # dsh-skin JSON CLI
│ ├─ mcp-server/ # MCP stdio 服务器(26 个工具)
│ └─ portable-runtime/ # 便携运行时聚合
├─ agents/codex-skill/deepseek-harness-skin-studio/
│ ├─ SKILL.md # Agent Skill 指令(自动换肤流程)
│ └─ scripts/ # open-studio / build-portable / install-local
├─ research/ # 同类项目调研材料(16 个项目 README)
├─ docx/ # 项目文档(实施计划 / 需求分析 / 架构设计)
└─ tests/ # 回归测试(含真实 rc.6 隔离 Chromium 预览)🛠️ Development
pnpm install # 安装依赖(需要 Node 22+ / pnpm 11+)
pnpm build # 构建全部包
pnpm test # 回归测试(会启动真实 rc.6 隔离预览)
pnpm controller # 本地启动 Controller(开发)
pnpm dev # Studio 开发模式(Vite)🗑️ Uninstallation
Remove all entries between
# >>> dsh-skin-studioand# <<< dsh-skin-studiocomment blocks in%DSH_HOME%\profiles\<profile>\cordis.patch.yml;Delete
%DSH_HOME%\profiles\node_modules\@dsh-skinand%DSH_HOME%\skills\deepseek-harness-skin-studio;Restart DSH;
To clean up Studio data, delete
%LOCALAPPDATA%\DeepSeekHarnessSkinStudio.
🔒 Security & Compatibility
Target DSH version:
0.1.0-rc.6(capability pinned; plugin fails closed if version mismatch).Skin model (ThemeSpec) only allows structured data: literal CSS colors, content-addressed images (sha256), controlled tokens; no arbitrary CSS/JS/URL.
Preview uses an isolated temporary DSH owned by the Controller; confirmation and writing must be done manually in the visible Studio.
Balance query: API Key is only parsed on the host side (
DSH_SKIN_BALANCE_API_KEY→DEEPSEEK_API_KEY→%DSH_HOME%\.credentials.yaml); the browser only sees aggregated balance, keys never land in the browser.
📄 Documentation
docx/— Implementation plan, requirements analysis specification, technical architecture design specificationCHANGELOG.md— Change logRELEASING.md— Release checklist (GitHub Release / npm)
📃 License
MIT © 2026 trrrrrryg
This server cannot be deployed
Maintenance
Related MCP Connectors
Remote MCP for RunComfy: ComfyUI deployments, hosted models, LoRA training. 31 tools.
Your org's AI agents, tasks, runs, search, and brain files as MCP tools and resources.
Free public MCP for AI agents — 193 tools, 44 workflows. No API key.
Generate on-brand images from your AI agent: design, edit, and render templates over MCP.
Related MCP Servers
- AlicenseBqualityCmaintenanceEnables AI agents to directly control Figma Desktop via MCP, supporting UI creation, editing, prototyping, and variable management with over 60 tools.657251MIT
- AlicenseAqualityCmaintenanceAugments DeepSeek-V4 with image understanding and deep research capabilities via MCP, enabling vision and web research tools.51MIT
- AlicenseNot gradedqualityBmaintenanceMCP server bridging ChatGPT to DeepSeek Harness, exposing 13 tools for task submission, status tracking, result retrieval, project/session management, and human-in-the-loop approvals.1MIT
- FlicenseAqualityCmaintenanceMCP server wrapping DeepSeek's agent harness tools into minimal/standard modes, enabling persistent shell, file editing, search, job management, and more via natural language.2-

