Skills Plugins MCP Prompt Model 导航 博客 资讯 我的中心
工具与能力 #agents-md#context-management#deepseek-harness#deepseek-harness-plugin#dsh#dsh-plugin

dsh-rich-context

Agent instruction manager for DSH — edit and template the AGENTS.md files the harness actually reads (global + per-workspace)

savageops @savageops ⬇ 2 ★ 0 main

安装

dsh plugin --profile web add github:savageops/dsh-rich-context
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。

Agent instruction manager for DSH — edit and template the AGENTS.md files the harness actually reads (global + per-workspace)

该插件未提供要点说明,请参考仓库 README。

agents-mdcontext-managementdeepseek-harnessdeepseek-harness-plugindshdsh-plugin
  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/savageops/dsh-rich-context
许可证MIT
主要语言main
下载量2
GitHub 星标0
最近推送2026-08-27
收录日期2026-09-19
分类工具与能力

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)。

# dsh-rich-context

**Agent manager for [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness)** — edit the `AGENTS.md` files your agents read, and define **subagent personas**: named roles with their own system prompt and a pinned `provider / model / effort` route, launchable as real sessions.

> 管理 AGENTS.md 与子代理人格:全局/工作区编辑、源扫描与默认切换;每个角色一个 Codex 风格 TOML(含 provider/model/effort 路由),编译为 DSH 预设、可从面板或模型的 agents 工具启动为真实会话。

## Install

```sh
dsh plugin --profile web add dsh-rich-context
```

Restart the `dsh web` process. An **Agents** entry appears in the sidebar (under Skill Center).

## Two modes

**Context** — the AGENTS.md manager (unchanged from v0.1):

- **Global** — edit `~/.dsh/AGENTS.md` with the monospace editor (Ctrl+S saves)
- **Workspace** — pick a workspace from the dropdown, edit its root `AGENTS.md`
- **Sources** — scan 10 tool directories, symlink any detected `AGENTS.md`/`CLAUDE.md` as the default

**Agents** — the persona roster:

- One `~/.dsh/agents/.toml` per agent: `name`, `description`, `provider`, `model`, `effort`, `sandbox_mode`, `developer_instructions` (the system prompt)
- Saves are free-form; the roster computes route health live against the model catalog; `agents launch` refuses invalid routes
- **Import** copies + converts foreign agent files (Codex TOML, Claude/Gemini markdown frontmatter), normalizing all five effort spellings; sources are never touched

## Model-facing `agents` tool

`agents launch` runs a persona as a one-shot **subagent** of the calling session: the persona file's contents become the child's system prompt, the child runs on the route its file pins, and its final output returns inline — the same contract as Claude Code / Codex subagents. `list` shows the roster, `read` one definition. Read-only personas get write/exec tools denied.

## Architecture

```
src/host.js            Node half — /api/rich-context/{state,file,template,sources,default}
                       plus /agents/{list,file,catalog,import,launch}: persona TOML store,
                       preset compiler, live model catalog, session spawner with seed
                       request/header route, and the `agents` tool registration.
src/client.bundle.js   Browser half — pure DOM overlay (no React dependency):
                       sidebar entry + panel with Context/Agents mode tabs.
```

## License

[MIT](LICENSE)

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。