Install
Inside DeepSeek Harness, with dsh-market
dsh plugin --profile web add dshmarket
Or from the command line
dsh plugin --profile web add dsh-ui-ux-pro-max
Installing runs third-party code with your own permissions — it can read your files, use your credentials and reach the network. Review the source first, and pin a commit (github:owner/repo#sha) when you can.
README
UI/UX design intelligence plugin for DeepSeek Harness (DSH), adapted and optimized from the open-source project nextlevelbuilder/ui-ux-pro-max-skill (MIT License). Bundles 84 design styles, 192 color palettes, 74 font pairings, 99 UX guidelines, 25 chart types across 22 technology stacks plus 1900+ Google Fonts records — fully offline, zero network, Chinese-first structured output.
中文说明见 README.md。
Tools
| Tool | Purpose |
|---|---|
design_recommend |
Product type + keywords → complete design system (ranked styles with reasons, HEX palette, font pairing, landing pattern, UX guidelines, anti-patterns, decision rules; optional stack + variance/motion/density dials) |
design_review |
UX checklist review of an existing UI, severity-ranked (Critical/High/Medium/Low) plus a pre-delivery checklist |
design_search |
Search the database by domain (style/color/chart/landing/product/ux/typography/icons/gsap/react/web/google-fonts) or by one of 22 stacks |
Chinese queries are auto-mapped to the English corpus (300+ term dictionary +
CJK bigram tokenization). A bundled ui-ux-pro-max skill teaches the model
when and how to use each tool. The plugin is host-only (no client UI).
Install
Plugin market / npm (recommended)
Published on npm as dsh-ui-ux-pro-max@1.0.3. Install from the DSH Desktop
plugin market, or with the CLI:
dsh plugin --profile desktop add --save-exact dsh-ui-ux-pro-max@1.0.3 # DSH Desktop
dsh plugin --profile web add --save-exact dsh-ui-ux-pro-max@1.0.3 # dsh web
💡 Note: DSH Desktop does not put the
dshcommand on PATH — open a new terminal (the desktop app ships its own host-commands); if the command is still missing, runnpm install -g @deepseek-ai/dshfirst.
Verified against DSH 0.1.1-rc.2 (
@deepseek-ai/dsh,dsh-tools,dsh-skill).
DSH Desktop (from source)
Copy this package into the profile's shared
node_modules:<profile>\node_modules\dsh-ui-ux-pro-max\(example:C:\Users\cwqjk\.dsh\profiles\desktop; packages live in the sibling...\profiles\node_modules\)Append to
<profile>\cordis.patch.yml:- insert: - id: ui-ux-pro-max name: dsh-ui-ux-pro-max config: {}Restart DSH Desktop.
Or run the idempotent installer (backs up the patch automatically):
powershell -ExecutionPolicy Bypass -File .\install.ps1
Web / CLI
Add the package to the profile's node_modules and either append the insert
entry above, or launch with pnpm dsh web --patch ./cordis.patch.yml.
Peer dependencies (@deepseek-ai/cordis, dsh-tools, dsh-skill) are
resolved from the host runtime — no extra npm installs.
Testing
No network and no DSH runtime required (CI runs these on Node 22 and 24):
node tests/check-package.mjs # manifest + bundled-data integrity
node tests/smoke.mjs # offline behaviour check (30 assertions)
check-package.mjs guards the failure modes actually hit in this project: a
UTF-8 BOM in package.json (breaks the DSH CLI's JSON.parse), a missing
dsh.bundle.patch, files[] entries that do not exist, lifecycle scripts
(rejected by the market), the cordis.patch.yml insert shape, and the
parseability/consistency of data/*.json.
smoke.mjs covers plugin behaviour: bundled database, design_recommend
(Chinese query → full design system), design_review (severity ordering),
design_search (domain/stack/Chinese mapping), and Chinese structured output.
Runtime compatibility is verified separately against a real DSH install (tool schema compilation, output schema,
render/presentCall, and theapply()registration path); that harness depends on local install paths and is therefore not shipped in this repository.
Credits
- Upstream: nextlevelbuilder/ui-ux-pro-max-skill (MIT) — data, search, and design-system logic ported and optimized for the DSH plugin form (pure JS, offline, Chinese-first, structured tool output).
- Plugin patterns inspired by
zhaiyateng/dsh-design-skills,Viger1/dsh-design, andsuperdesigndev/superdesign-skill.
License
MIT © ChenYiming-aaa. Third-party data/algorithm sources are listed in THIRD_PARTY_NOTICES.md.
Comments
Comments live in GitHub Discussions. Sign in with GitHub to post or react.