생산성 도구 · DeepSeek Harness 플러그인 — 189 페이지

让 AI 在开发交流中优先用 Mermaid 图表达：全局 prompt 引导 + mermaid_validate 语法校验。渲染交给 dsh-mermaid（MrmoLabs）。

DeepSeek Harness Web UI plugin that renders settled Mermaid fenced code blocks as diagrams

DSH Web 用户消息编辑重发：把某一轮提问改掉重发，该轮及其之后的内容丢弃（原会话归档，新会话继承逐字节相同的前缀）。

Per-message session handoff for DeepSeek Harness: a handoff icon next to each message's fork button summarizes the conversation up to that message with an LLM, forks a fresh session carrying the copied history plus the injected summary document, and auto-opens it. The origin session is unchanged.

DSH (DeepSeek Harness) web 消息增强插件：① 会话区左侧/右侧丝滑悬浮指令导航条（Codex 风格，可配置位置），刻度覆盖本地会话完整历史（含未加载部分，点击即加载更早并定位），滚动联动定位、悬停预览指令内容；② 输入框 ↑/↓ 键历史消息穿梭，像终端命令历史一样把已发送内容快速回填到输入框。

为 DSH 移植 Codex 风格的会话消息导航轨道和悬停摘要

Sidebar message-history preview for DeepSeek Harness web UI: lists the current session's sent messages with previews; click to jump the chat to that message.

删掉不喜欢的 AI 回复，或删掉已发送的错误提示词 —— 每条消息可撤回 / 删除 / 删除此处及之后。DSH 插件：写 append-only 墓碑，模型上下文与界面同步移除。Per-message recall & delete for DeepSeek Harness sessions.

Codex 风格消息原地编辑（DeepSeek Harness / DSH 插件）：hover 历史用户消息 → ✎ 编辑 → 同会话从编辑点截断并重新生成后续回复。

The DeepSeek tariff meter for dsh: what this session cost, which time-of-use tariff you are standing in, and when it changes — one line under the composer.

mi logo + quota ring beside the model selector, plus a MiMo usage detail tab (billing type, plan quota, per-session attribution, cost estimate).

MiMoCode-style Zero-tool Workspace Memory Auto-Injection & Background Distillation Plugin for DeepSeek Harness

Realtime model usage panel: tokens, cost, requests, trend and composition

mi logo + quota ring beside the model selector, plus a MiMo usage detail tab (billing type, plan quota, per-session attribution, cost estimate).

DeepSeek Harness (dsh) plugin: prune the tool schema and inject a structural preamble on a session's first turn only.

Worker orchestration for DeepSeek Harness: session_create / delegate_task / persona_list / tool-nudge

ML 管线工作流：数据→训练→评估→部署→监控的 MLOps 工程。受 wshobson/agents（38k★ MIT）启发。

Mnemosyne memory for DeepSeek Harness: remember / recall / forget / stats / sleep tools, embedded skill, auto-setup, optional auto-sync/prefetch, and a Settings panel. Data lives under ~/.dsh/mnemosyne.

Android Mobile GUI Agent plugin for DeepSeek Harness with ADB control, verified iterative actions, approval, and a Web mobile view

Mobile experience suite for the DeepSeek Harness web UI — reading layout (<1024px drawer sidebar + whale button), aurora glass skinning with auto text contrast, file-browser tab, sidebar file upload to the workspace, abort auto-heal. Zero dependencies.

DeepSeek Harness WeChat remote-control plugin: control your dsh agent from WeChat (iLink long-polling, fully outbound, works over any network)

Mobile UI overlay for the DeepSeek Harness Web GUI: bottom navigation, session drawer, and small-screen panel orchestration as an out-of-tree client plugin (no upstream source changes).

模型无关的沙箱升级修正 + 按模型定制的协议提示段：任何模型（GPT 系最常见）在工具调用中错误携带 sandbox_permissions/justification 导致 "not strictly wider" 死循环时，在工具执行前自动剥离；并按会话模型动态注入适配该模型的协议提示段。

DSH 模型健康检查：在会话视图新增「模型健康检查」页签，按「供应商(baseURL分组) → API类型 → 模型 → 路由」四行按钮选定目标，手动发送一条真实裸 HTTP 测试请求（支持 openai-completions / openai-responses / anthropic-messages 三协议，流式与非流式可切），一屏展示耗时、TTFT、HTTP 状态、token 用量、响应内容与三层诊断链（API请求 / 模型响应 / 严格校验），错误全部人话化。可把当前模型一键设为默认（下次打开即预选）。另提供用户显式启动、成功即停的有限自动重试会话（严格/连通两种成功判定、4 条停止条件、致命错误不短路）。不做后台监控、不轮询、不告警、不落盘。