Ship a local DSH plugin project to GitHub and into the plugin market in one call: preflight checks, per-file upload with byte verification, the market registry PR, then pull every published file back and compare bytes.
생산성 도구
$ npx -p @deepseek-ai/dsh dsh plugin --profile web add github:7gzgn/dsh-slipway
DSH web plugin: arrange the list slots other plugins contribute to — show/hide, reorder, rename — without touching upstream. Ships a slot-agnostic engine (three SlotRegistry.prototype patches) driven by a small managed-slot table; today it manages settings.section, sidebar.panellist and sidebar.footer.action.
인터페이스 확장
$ npx -p @deepseek-ai/dsh dsh plugin --profile web add github:Yiklek/dsh-slots-manager
Custom DSH CompactionEngine backend: configurable threshold prompting, mandatory fallback at the limit, DCP-inspired dedup/purge, custom summary prompt, and /compact like the original. Zero external state — everything is rebuilt from the append-only session log.
Smart calendar reminder & schedule assistant for DSH Web GUI: lunar calendar, holidays, system notifications on macOS/Windows, and WeCom/platform push integration.
알림 · 서비스 연동
$ npx -p @deepseek-ai/dsh dsh plugin --profile web add github:a792883583/dsh-smart-reminder
Automatic model routing for DeepSeek Harness: a virtual `smart` model classifies every request by difficulty (hard / normal / easy) and by vision need, then delegates to the models you already configured in Settings → Models. 内置智能路由:三级难度 + 视觉自动路由,模型从「设置-模型」已配置模型中选取,虚拟 smart 模型零配置接入。
시각 도구
$ npx -p @deepseek-ai/dsh dsh plugin --profile web add github:rouyiemei/dsh-smart-router
DSH Web plugin: a Smart Prompt button under the composer that rewrites your draft into a structured prompt using your workspace conventions, then fills it back into the composer.
Agent 협업 · 생산성 도구
$ npx -p @deepseek-ai/dsh dsh plugin --profile web add github:shellpower96/dsh-prompt-optimizer
Two DeepSeek Harness agent presets: `dsh-smith`, which builds harness agents and Cordis plugins, and `dsh-forge`, a software-delivery agent. Each carries a thinking protocol, a layered memory discipline, and a standing team of named experts.