DeepSeek Harness Plugin Hub

发布与管理完整 Harness Profiles，发现适合你的插件。

探索

插件目录环境预设文档中心动态

社区

发布插件联系我们报告问题

相关链接

Plugin Hub GitHubDeepSeek Harness 官方项目系统状态隐私说明
© 2026 DeepSeek Harness Plugin HubPowered byPaxTech

独立、非官方社区项目，与 DeepSeek 官方无隶属、授权或背书关系。

Prompt Lib — DeepSeek Harness 插件（DSH Plugin）
← Plugins
P

@stmol/dsh-prompt-lib

Prompt Lib

DeepSeek Harness 的用户自定义提示词库：在设置中提供“提示词”页面，以及一个编辑器选择器，可将所选提示词注入会话及其子代理

插件会安装到这里；不确定时保持 web。

npx -y @deepseek-ai/dsh plugin --profile web add github:Stmol/dsh-prompt-lib#fc6dd22fd2855b18dd55cd01238680a76b20bc8d
README兼容性版本
Settings → Prompts page with a prompt editor open and a saved prompt below

兼容性与来源证明

Prompt Lib 以 @stmol/dsh-prompt-lib 发布，当前版本为 1.0.1。Plugin Hub 会校验它的 manifest，并保存精确安装来源，便于复现安装结果。

DSH 兼容范围
*
运行环境
web
发布来源
github
Registry 更新时间
2026/10/4

版本

1.0.1stable
2026/10/4
1.0.0stable
2026/10/4

相关插件

正在加载相关插件…

最新版
1.0.1
DSH
*
HMR
重启进程
Tree shaking
未声明可安全裁剪
解包体积
未提供
文件数
未提供
Surface
web
许可证
MIT
发布源
github
GitHub
★ 0
周下载
0
最近提交
2026/10/4
查看源码 ↗
README Badge

点击下方 Badge 复制 Markdown，粘贴到 README 即可。

这是你的 Plugin？认领权益 · 优先安全扫描

验证 package.json 声明的 GitHub 仓库，即可管理这个公开页面。认领后，Hub 会优先安排当前版本的安全扫描，并在通过后公开展示结果。

认领这个 Plugin →
报告问题
DeepSeek Harness Plugin Hub
ProfilesPlugins分类动态文档登录管理 Profiles
ProfilesPlugins分类动态文档登录

相关插件

继续浏览 memory-context 分类下经过校验的插件。

Memory Plugin@openviking/dsh-memory-plugin适用于 DeepSeek Harness 的 OpenViking 记忆与上下文套件Contextdsh-context用于上下文洞察和管理的 DeepSeek Harness 插件，提供上下文仪表板和上下文命令，帮助了解上下文的构成及其演变过程。Weknora@wxg-prc-cpg/dsh-weknora适用于 DeepSeek Harness (dsh) 的 WeKnora 知识检索工具：通过自有知识库进行语义搜索、文档阅读以及 RAG/代理回答。Memsearch Dsh@zilliz/memsearch-dsh适用于 DeepSeek Harness 的 MemSearch 插件：在多个代理之间共享 Markdown 记忆，支持捕获、步骤前上下文注入、记忆召回技能和技能候选审核面板。

README

@stmol/dsh-prompt-lib

🥷 Keep your own prompts inside DeepSeek Harness, and switch them on for a chat in one click.

  • Settings → Prompts: write, edit and delete your prompts.
  • Chat composer: tick the ones you want for this session. The button shows how many are on.

A prompt that is ticked applies to every model step of that session, and to the subagents it starts.

Prompts picker open in the chat composer, one prompt ticked and one available

Settings → Prompts page with a prompt editor open and a saved prompt below

Install

Published on npm as @stmol/dsh-prompt-lib. Any bundle source works: the npm package, the repository, a tarball, or a local folder.

  • GUI: Settings → Plugins → Add plugin, then @stmol/dsh-prompt-lib.
  • CLI: dsh plugin add @stmol/dsh-prompt-lib.

Restart DSH after the first install; later edits reach the running app through HMR. There are no automatic updates: remove the plugin and add it again.

Install from a local checkout 
dsh plugin add link:/path/to/dsh-prompt-lib

A linked checkout keeps its own node_modules, so the plugin's one dependency has to be installed there too:

cd /path/to/dsh-prompt-lib && npm install

Without it the plugin fails to load and nothing appears in Settings.

Usage

  1. Open Settings → Prompts. It is its own page, after Agent presets. Write a title and an instruction, then save.
    • The title is a label for you. The list, the search field and the composer picker use it to keep prompts findable. It is never sent to the model, only the instruction text is.
    • The instruction is stored in the plugin's config, so it survives restarts and is there in every session.
    • Enable in every new chat marks the prompt as a default. It is off unless you turn it on.
  2. In a chat, open the Prompts control in the composer tool row, right of the permissions dropdown. Tick what you want. The button shows the count.
  3. The ticks apply from the next model step. No restart and no new session needed. Tick again, or press Clear, to take them out.

Prompts are matched by an id generated behind the scenes, not by title, so renaming a prompt keeps every session pointed at it.

Defaults for new chats

A prompt can be set to start enabled. Every session that has never been ticked by hand opens with those defaults already on, and the picker shows them ticked, so the count on a fresh chat matches what that chat will use.

The first hand edit to a session, whether that is ticking, unticking or Clear, becomes that session's own choice and takes over from then on. Unticking the last default prompt records "none here" instead of falling back to the defaults you just removed.

Subagents

A session that has ticks of its own uses exactly those, and an empty list counts as a choice. A session with no choice at all takes the choice of the nearest ancestor session, so a subagent inherits from the chat that started it. Only a whole line with no choice anywhere falls back to the defaults for new chats.

Good to know

  • The settings file must be writable. The Prompts page says so up front and disables its controls. In the composer, a write the host refuses shows the reason under the list, so an edit is never dropped silently.
  • The composer control hides itself until the settings remote is mounted and has answered its first read, rather than show a count that is about to change.
  • There are size limits. The editor caps one instruction at 8000 characters. The host applies that same cap, uses at most 100 prompts and reads at most 500 library entries when it renders. Anything past those limits contributes nothing to the model. Settings still shows, edits and keeps every entry.
  • Selections accumulate. The plugin keeps one entry per session that has ever been ticked in, and never prunes them. An entry whose prompts were all deleted still means "inject nothing", so removing it would quietly bring back the defaults for that session.