A DeepSeek Harness–native plugin marketplace (installed via the official `dsh plugin` command, runs zero-token passively in the DSH sidebar; Web edition at dsh.market). License recorded as NOASSERTION on GitHub while the README declares MIT; actively maintained.
DSH integration
Native runtime
Author-claimed
Safety audit
Unaudited
Last verified
2026-08-28
License
NOASSERTION
01What can it help you accomplish?
Find and evaluate DeepSeek Harness plugins with Chinese search and practical scoring
a continuously refreshed catalog (currently 5,861 plugins + bundle channel) with Chinese intros, 5-dimension scores, radar charts and real install commands
DeepSeek Harness users who want to discover, compare and pick plugins before installing
Install, detect and remove DSH plugins from the sidebar
one-click install (skill-type via `git clone`, cordis-type via `dsh plugin add`) with auto-retry and rollback; installed detection and one-click uninstall via the Installed (已装) tab
DSH users, including beginners — the cold-start questionnaire delivers curated picks within about 3 minutes
Hand the install of uncertain plugins to an AI agent
plugin installed with minimal LLM usage — T0 zero-LLM direct install first (recipe cache / parsed commands + smoke verification; 75%+ plugins need no AI), minimal-protocol subagent only when needed; recipes auto-learned and install metrics recorded locally
Users who are unsure whether or how to install a plugin
02How to install into DeepSeek Harness
Prerequisites
- Web edition: zero install — just open https://dsh.market/ in a browser
- Sidebar plugin: a DeepSeek Harness environment — it is added through the official `dsh plugin` command and requires a harness restart afterwards
Installation steps
- 01
Web edition: visit https://dsh.market/ directly (no installation)
- 02
Sidebar plugin: run `npx @deepseek-ai/dsh plugin --profile web add @dsh-market/plugin`
$ npx @deepseek-ai/dsh plugin --profile web add @dsh-market/plugin
- 03
Restart the harness — a 插件市场 (Plugin Market) entry appears at the bottom of the sidebar
Verify the integration
- After restarting the harness, the 插件市场 (Plugin Market) entry appears at the bottom of the sidebar
Rollback
- One-click install failures auto-retry and can be rolled back; the Installed (已装) tab offers one-click uninstall for installed plugins
03DSH integration and capability boundaries
First-class dsh plugin — installed via the official `dsh plugin` command (`npx @deepseek-ai/dsh plugin --profile web add @dsh-market/plugin`) and running as a cordis plugin in the DSH sidebar (zero-token passive); the Web edition at dsh.market shares the same plugins.json
One-click install
any plugin in the catalog — skill-type or cordis-type→plugin installed on the machine: skill-type via `git clone`, cordis-type via `dsh plugin add`; auto-retry and rollback on failure
Installs code locally — skill plugins are git-cloned, cordis plugins are registered via `dsh plugin add`Practical 5-dimension scoring
plugin metadata — recent commits & issue health, README/docs/examples completeness, stars & fork participation, install clarity, license/topics/description/README completeness→a weighted geometric-mean score (maintenance 30% / practicality 25% / ecosystem heat 20% / convenience 15% / signal quality 10%) plus a per-plugin 'why recommended' explanation
AI agent install (route mode)
any plugin the user wants installed, including ones they are unsure about→installed plugin with minimal LLM usage: T0 zero-LLM direct install (installed check → recipe cache → parsed README commands + smoke verification; 75%+ plugins need no AI), minimal-protocol subagent only when needed, JSON verdicts auto-learned as recipes
Installs code locallyInstall metrics (T0 hit rate / AI participation / success rate) are recorded locallyAsks for confirmation before applying configurationRecommendation system
cold-start questionnaire, favorites, GitHub stars, installed plugins; scenario mode reads current session title & messages (manual trigger)→personalized 'guess you like' picks with EMA decay and scenario recommendations, each with a 'why recommend' reason
User profile (questionnaire / favorites / GitHub stars / installed plugins) stays on the local machineScenario recommendation reads the current session title and messages (zero token)Integration pack (整合包) ecosystem
dsh.pack.json bundle manifests→browse and install integration packs — Web 整合包 section + plugin-side 整合包 Tab (browse / entry parse-rate validation / install entry)
04Who is it for? When not to use it?
Good for
- DeepSeek Harness users who want to discover, compare and pick plugins before installing
- DSH users, including beginners — the cold-start questionnaire delivers curated picks within about 3 minutes
- Users who are unsure whether or how to install a plugin
05Compatibility, maintenance and safety notes
- Only DeepSeek Harness ecosystem plugins are indexed — general-purpose projects are explicitly excluded, and a repo must pass feature detection (SKILL.md or cordis markers) even if it carries the dsh-plugin topic.
- Catalog data is refreshed by a daily 06:00 automated pipeline (stars and descriptions refresh every day at 06:00), so scores and metadata are as current as the latest pipeline run.
- AI semantic search is announced as 'coming soon' and is default-off to save tokens: local recall of 60 candidates → LLM re-ranks 20 with reasons, with a fixed candidate pool that does not grow with plugin count.
NOASSERTION (README declares MIT) · active, last push 2026-08-26, no releases
06Frequently asked questions
How do I install the sidebar plugin into DeepSeek Harness?
Run `npx @deepseek-ai/dsh plugin --profile web add @dsh-market/plugin`, then restart the harness — the sidebar gains a 插件市场 (Plugin Market) entry at the bottom. The Web edition needs no installation: just open https://dsh.market/.
What is the difference between the Web edition and the plugin edition?
The Web edition is a browser-based GitHub Pages static site (zero install) aimed at discovery and evaluation; the DSH plugin edition is a cordis plugin living in the DSH sidebar and handling install and management (5-tab panel: Recommend / Search / Favorites / Installed / Settings). They only share the same plugins.json data, refreshed daily at 06:00, and are otherwise independent.
How does one-click install decide how a plugin gets installed?
Deterministic routing: skill-type plugins are installed via `git clone`, cordis-type plugins via `dsh plugin add`; on failure it auto-retries and can roll back. The Installed (已装) tab detects what is installed on the machine and offers one-click uninstall.
Does the AI install feature consume tokens?
Route mode prioritizes zero-LLM direct install (already-installed detection → recipe cache → parsed README commands + smoke verification; 75%+ plugins need no AI) and only delegates to a minimal-protocol subagent when needed. The plugin itself runs passively at zero token cost — opening the panel is the only cost.
What plugins does the market actually collect?
Only DeepSeek Harness ecosystem plugins: the daily 06:00 pipeline scans GitHub topics (dsh-plugin / dsh / deepseek-harness-plugin / dsh-bundle / dsh-skill), awesome lists, the dsh-external org and submitted issues, and a repo must pass feature detection (SKILL.md or cordis markers) to be listed — currently 5,861 plugins plus a bundle channel.
07Related DSH workflows
dsh-context
by bowenliang123
Context dashboard tab + /context command showing context-window composition, per-turn history, compaction/prune events, per-message token costs
api-relay-audit
by toby-bridges
Local security audit for AI API relays and LLM proxies: detects prompt injection, model substitution, tool-call rewriting, SSE anomalies, error leakage, and Web3 wallet risks.
jingyun-dsh
by jingyunstudio
A one-stop AI commercialization desktop client built on Jingyun Studio plus DeepSeek Harness (DSH).
awesome-deepseek-harness-plugins
by zhiyuan-fan
Curated DeepSeek Harness (DSH) plugins, extensions, tools, skills, clients, runtimes, integrations, and verified references — English and Chinese.
08Data and sources
DSH 侧边栏 · cordis 插件
`npx @deepseek-ai/dsh plugin --profile web add @dsh-market/plugin`
装完**重启 harness**，侧边栏底部出现「插件市场」入口。
This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-08-28. Found an error? Submit a correction.
Best DeepSeek Harness Plugins
Twelve plugins worth installing first — picked from the whole catalog, across every category.