스킬 · DeepSeek Harness 플러그인 — 27 페이지

统一管理 DSH 插件/Skill/MCP，支持角色化动态装载与切换

Host-neutral durable DAG workflows for Agents, Tools, Skills, MCP, triggers, replay, and visual Canvas

Durable cross-session memory and self-evolving skills for the DeepSeek Harness

Agent Skill for DeepSeek Harness (dsh): biga.

wxpilot — A CLI for automating WeChat Mini Programs, built for AI Agents. Lets an Agent drive the WeChat DevTools like a browser — page navigation, element interaction, state reading, network capture & mocking. 面向 AI Agent 的微信小程序自动化 CLI 让 Agent 像操作浏览器一样操作微信开发者工具——页面导航、元素交互、状态读取、网络抓包与 mock。

Zero-signup public plaza for AI agents — speak, read, discover

Deploy the ArchGraph ARGO toolchain, skills, and rules (schema, scripts, argo-init skill, global rule) with one command.

面向 AI Agent 的 Markdown 转 Word 工具，提供开箱即用的中文排版、Mermaid 图表转图片和可编辑数学公式，提供Skill、独立 CLI 和 DSH 插件。

DSH plugin adapter of catalog-capabilities-zh: curated Chinese capability catalog; orchestrates DeepSeek Harness plugins/skills installation and documentation from user-selected or manually provided sources.

冷咖啡执行核：一个插件，三档破甲强度，靠配置文件切换（lite / mid / deep）。86 条三拍技能 + 关键词路由。

AtomGit plugin bundle for DeepSeek Harness: atomgit-skills workflows + ag CLI integration + GitCode MCP server tools

Agent Skills and Rules Workbench for DeepSeek Harness better-sidebar

A DSH Bundle that registers a multi-source capability discovery and risk inspection Skill.

Content-addressed receipts for skills actually loaded by DeepSeek Harness

DSH 对话体验增强套件 + 四工作台（零 DSH 源码改动，纯插件注入）。回合呈现：思考 / 工具调用聚合 chip（实时走秒 + 分步跟随滚动）+ 步骤卡 / 总结卡（回合收口才出现）+ 共享活动抽屉（思考语义分组 + 工具调用树）。正文增强：proto-tabs 可交互卡片 + diagram 流程图围栏（JSON 转 SVG）+ 生图画廊 + 重试行影子。界面与工具：会话头部视图标签上移右上角 + 桌面壳窗口控制留位与主题同步 + 对话截图（常驻无头浏览器渲染 markdown/shiki/mermaid，可内嵌本地 HTML）+ download 下载工具（wire 工具 + 实时进度条/速度/ETA）+ KR 对话双栏大盘（右栏保留任务 / 思考 / 工具调用 / 记忆四张卡，顶栏由 KR_PANEL_HEADER_VISIBLE 门控）。融合的原 dsh-triad 四工作台（座位与路由前缀原样保留，dsh-triad 自此退役）：自动沉淀的长期记忆、定时自动化、用量与账号趋势、技能与 MCP Server 管理。

Vision-only desktop automation plugin for DeepSeek Harness (fusion refactor)

Data cleaning and data enrichment for CSV/XLSX/JSON enterprise lists in DeepSeek Harness, including spreadsheet cleaning, deduplication, profiling, optional Qichacha MCP and exports.

DSH (DeepSeek Harness) data-insight skill plugin: turn raw data (CSV / pasted table / SQL results / DuckDB) into a structured Markdown report with business conclusions, metrics and charts.

Diagrams (row `dsh-diagrams`, tab kinds `diagram` / `diagrams`): Mermaid and TikZ LaTeX diagrams as a first-class vncode surface. The model gets six tools (diagram_write/patch/read/verify/publish/delete), each taking an optional conversation-or-library scope, that validate every diagram on the host - Mermaid is parsed headlessly by the same vendored engine the browser renders with, TikZ is compiled by the machine's own TeX engine - and the verdict comes back with line-accurate diagnostics plus advisory warnings about what will not read well. Nothing the engine builds can reach the page: mermaid renders behind parse-first validation with its own error drawing suppressed and a container the plugin owns, so a broken source is text and never a 'Syntax error in text' picture. The browser reports back what it actually drew, so the model can verify a picture exists rather than just that the source parses. Plus two bundled skills that teach Mermaid and TikZ craft. Each diagram becomes a real right-bar tab with a rendered view laid out at 80% of the pane, a 25%-400% zoom ladder that moves the layout box (so a zoomed diagram stays scrollable to its edge), a source drawer and an export menu whose every format - mmd/md/tex/pdf/svg/png - is SAVED TO THE DESKTOP of the machine running the harness by the host (create-exclusively, never the conversation folder, with the browser download kept as the fallback), and every tool call renders the diagram inline in the conversation. The browser's verdict on a revision is ONE function with four objective states (drawn / failed / stale / pending, carrying both revision numbers) shared by the tool result, the index and the tab pill, and a cache hit re-reads the verdict the compile was cached with rather than assuming success. Each bundled skill also ships a reference/complex-diagrams.md for pictures too big for the syntax summary. State is per-conversation plus one harness-wide library, and survives restarts; artifacts are content-addressed and cached. No core patches, no fork, no npm dependencies, no network. Alpha.

EPLAN Electric P8 integration for DeepSeek Harness: a bundled local EPLAN automation MCP server, an offline P8 2025 .NET API reference and platform guide built from the installed docs, and offline 2026 / 2027 / EEC Pro documentation lookups (those three ship disabled and read indexes extracted from the release assets).

DeepSeek Harness 插件：把 Codex 技能 fight-scene-director（多画风 AI 视频打戏设计与轨迹图调度）移植为 DSH 原生技能，输出适配 Seedance 2.0 / Seedance 2.5 / MiniMax H3 的结构化中文提示词。

Bridge the vercel-labs/skills ecosystem into dsh: LLM-driven skill search, install, and lifecycle for temp/project/global scopes

DeepSeek Harness plugin: 鱼排 FishPai — a WeChat Official Account typesetting workbench in the DSH right sidebar, with block-level diffs and annotations a model can read. Rendering core derived from the MIT mopai-markdown.

Three learning skills for DeepSeek Harness: course notes, conceptual explanations, and self-contained lecture chapters.