Skills Plugins MCP Prompt Model 导航 博客 资讯 我的中心
界面增强 #deepseek-harness#design#dsh-plugin#claude-design#dsh#dsh-bundle

deep-design

Design mode for DeepSeek Harness: the design-loop agent preset plus design-principles and design-qa skill packs

temidayoxyz @temidayoxyz ⬇ 1 ★ 0 master

安装

dsh plugin --profile web add github:temidayoxyz/deep-design
下载安装清单

需要可复现安装时,可在仓库后追加 #commit 固定提交。

Design mode for DeepSeek Harness: the design-loop agent preset plus design-principles and design-qa skill packs

该插件未提供要点说明,请参考仓库 README。

deepseek-harnessdesigndsh-pluginclaude-designdshdsh-bundle
  1. 安装并启动 DeepSeek Harness:npx @deepseek-ai/dsh web
  2. 在终端执行上面的安装命令(CLI 会解析插件并核验来源)
  3. 用 dsh plugins list 确认已安装,必要时重启 Harness 生效

插件以当前 dsh 进程的权限运行,安装时可能执行代码。请先通读仓库源码与许可证,确认无破坏性命令与越权访问;本站只做索引,不对第三方插件安全性作担保。

代码仓库github.com/temidayoxyz/deep-design
许可证MIT
主要语言master
下载量1
GitHub 星标0
最近推送2026-08-15
收录日期2026-09-19
分类界面增强

事实信息来自公开插件目录快照(2026-10-01),介绍文案由本站再加工。

以下为插件仓库 README 全文(原始内容,由公开目录抓取整理)。

# Deep Design

[图片: Deep Design — the render glass]

**Design mode for DeepSeek Harness** — a self-hosted alternative to Claude Design, Open Design, Google Stitch, and Figma Make, built on the DeepSeek Harness agent engine.

Deep Design turns your local harness into a design studio. It runs the same loop those products run — brief → direction → artifact → critique → deliver — with rendered visual QA at every pass, on your machine, with your own models and API keys.

## What it installs

One plugin, three things:

1. **Deep Design** — an agent preset in your roster: a complete design agent (shell, files, web research, skills, plans, subagents, workflows) whose entire job is the design loop.
2. **`deep-design-principles`** — the design intelligence: first-principles brief, design language before implementation, composition, motion, anti-slop audits, a scoring rubric, and a final self-review.
3. **`deep-design-qa`** — the visual feedback loop: headless render at desktop and mobile → a critic pass (a subagent that diagnoses but never writes code) → targeted fixes → re-render → compare. With a pixel-sampling fallback for models without image input.

Both skills register globally, so any agent mode can load them from the skill catalog — not just Deep Design sessions.

## Install

```sh
dsh plugin --profile web add dsh-deep-design    # from the npm registry once published, or
dsh plugin --profile web add ./deep-design      # from a checkout
```

Restart the Web UI, open the new-session screen, and choose **Deep Design** from the agent-preset chip. (The `web` profile is the one `dsh web` boots; use the same command with another profile name for other surfaces.)

The plugin installs the preset into `$DSH_HOME/.agent-presets/design/`, the standard user-preset directory every deployment scans. Installation is idempotent and never overwrites an existing directory — once you edit your copy, it is yours.

## The loop

```
brief → direction (.design/direction.md) → design system → artifact
     → render (headless, desktop + mobile)
     → critic pass (structured visual diagnosis)
     → fix → re-render → compare → deliver
```

Design memory compounds across sessions: the loop reads `.design/` in the workspace at session start and records lessons at session end. Designs get better with use, not just with better models.

## Notes

- Best with a vision-capable model for Deep Design sessions — the critic sees better with eyes. The pixel-sampling fallback keeps the loop honest without one.
- Web research is enabled on the preset (`fetch: true`); configure the harness's web provider for reference grounding.
- Deep Design runs entirely on your machine. Nothing leaves it unless you choose a cloud model.

## License

MIT.

数据来源:公开的 DeepSeek Harness 插件目录与各插件 GitHub 仓库。本站为独立第三方目录,与 DeepSeek、幻方(High-Flyer)及插件作者均无隶属或背书关系。

每日精选 Skill 推荐,免费送到你邮箱

输入邮箱,每天接收一个精选 AI Agent 技能推荐。完全免费,持续更新。

提交后我们会发送一封确认邮件,点击邮件里的链接才会开始收信。

完全免费,取消任意时间。我们不会发送垃圾邮件。