Skill Deck

io.github.xingbofengPublished by GitHub account xingbofengv1.0.0Updated Oct 24, 2026

Expose local Agent Skills as adapter tools, MCP resources, MCP tools, and share artifacts.

VerifiedSTDIODesktop onlyProductivity & WorkflowKnowledge & MemoryDeveloper Tools

Overview

AI-generated overview

SkillDeck exposes local Agent Skills as MCP tools and resources so an assistant can discover, search, load, and share them.

What you can do
SkillDeck scans, parses, and validates local SKILL.md files and exposes them to an MCP host as tools and resources. It offers three activation modes: compact (list, search, info, read, resources), guided (adds a skill_guide and MCP instructions), and active (adds up to N use_skill_* tools so common skills appear directly in tools/list). It also generates share artifacts through agent conversation, including an index.html page plus cover and detail images, with local paths, secrets, and private URLs removed by default.
When to use it
Use it when you keep a set of Agent Skills on your machine and want an MCP host such as Claude Code or Cursor to discover and load them during tasks. It is also useful when you want to hand a sanitized, browsable summary of those skills to colleagues or peers.
Requirements
Runs locally over stdio, typically via npx skill-deck mcp serve. Requires Node.js and the npm package skill-deck. The environment variable SKILL_DECK_SKILLS_ROOT is required and must be an absolute path to the local skills root directory. Desktop only; no remote endpoint or authentication is declared.
Before you install
The server reads the local skills directory you point it at, so scope SKILL_DECK_SKILLS_ROOT carefully. Share generation writes files such as index.html, cover.png, detail.png, and manifest.json to an output directory you specify. Sharing is described as sanitized by default, but review generated artifacts before sending them to others, since skills may contain local paths, secrets, or private URLs.

Installation

In SourceWeft

  1. Open Skill Deck in the dashboard and add it to a workspace.
  2. Enable the server for the chats that should use its tools.

Desktop only via STDIO. STDIO servers start a local process, so they need the SourceWeft desktop host.

Other MCP clients

Follow the launch instructions in the repository.

README

SkillDeck

[SkillDeck logo]

中文 · English · 官网 · 快速开始 · GitHub

把散落在本机的 Agent Skills，变成模型看得见、用得上、也分享得出去的能力牌组。
Adapter for runtimes, MCP dynamic discovery, and share artifacts generated through agent conversation.

[Runtime Node.js 20+] [Package skill-deck] [License MIT] [MCP stdio resources tools] [Skill Activation active by default] [Share HTML and Images] [OpenAI Agents SDK] [Claude Code ready]

[SkillDeck: Adapter, MCP, Share]

SkillDeck 是本地 Agent Skill 的适配、激活与分享工具。

它解决三个很具体的场景：

场景为什么需要它SkillDeck 做什么
我有一组 Skill，想通过工具的方式接入 OpenAI Agent SDK。OpenAI 已把 SKILL.md 定义成可复用工作流，Agents SDK 也支持 local skills，但手写 schema、handler 和 stable id 很繁琐。Adapter 扫描、解析、校验 SKILL.md，生成 provider-neutral tools 和 runtime handlers，让同一套 Skills 可以接入不同 Agent runtime。
我有一组 Skill，想接到 Claude Code、Cursor 或其他 MCP host。只给 list_skills / read_skill 不够，模型不一定知道复杂任务要先查本地 Skills。MCP + Skill Activation 把 Skills 暴露成 tools/resources，并通过 compact / guided / active 三种模式，让模型在任务里自己发现、搜索和加载 Skill。
我有一组 Skill，想发给同事、同行或社交平台看。私有 Skill 往往含本地路径、secret、私有 URL；只发截图又无法搜索、下钻和复用。Share 通过 MCP 让 agent 生成完整分享页和双图，默认脱敏，可搜索、可下钻、可审阅再接入。

背景参考：OpenAI Skills guide、Skills in ChatGPT、Agents SDK tools、openai-agents-python #2906。

对应到产品能力，它做三件事：

主线作用结果
Adapter扫描、解析、校验本地 SKILL.md，生成 stable id、provider-neutral tools 和 runtime handlers。同一套 Skills 可接 OpenAI、Anthropic、OpenAI Agents SDK、本地 runtime。
MCP把 Skills 暴露成 MCP tools/resources，并通过 Skill Activation 帮模型动态发现。模型可以通过 skill_guide、use_skill_*、search_skills、read_skill 加载 Skill。
Share你和 agent 说“帮我分享这套 Skills”，agent 通过 MCP 生成静态分享页和双图，默认脱敏。适合朋友圈、小红书和同行交流，不暴露本地路径、secret 和私有 URL。

它不是公开 marketplace，也不是远程托管服务。用户的 Skill 目录仍然在本机。

Skill Activation

默认的 list_skills / read_skill 能读 Skill，但模型不一定会主动调用。SkillDeck 增加三种模式，让模型更容易发现并加载本地 Skill。

模式暴露内容模型怎么发现 Skill适合场景代价
compactlist_skills、search_skills、get_skill_info、read_skill、resources模型需要主动搜索或列目录兼容旧行为、Skill 很多、工具列表要短自动发现能力弱
guidedcompact + skill_guide + MCP instructions模型看到 skill_guide 后获得使用指南想提高发现概率，但不想每个 Skill 都变 tool仍依赖模型调用向导
activeguided + 最多 N 个 use_skill_<safe_name>_<short_id>模型在 tools/list 阶段直接看到常用 Skill想最大化自动发现效果工具数量增加

示例：

bash
npx -y skill-deck mcp serve \  --skills ~/.skills

安装

安装：

bash
npm install skill-deck

MCP 三步接入

第一步，选择一种 Skill Activation 模式安装 MCP。推荐先用默认 active：

三种命令的 MCP server 名都叫 skill-deck；切换模式时更新同名 server 即可。

bash
# active：默认模式，常用 Skill 会直接出现在 tools/listclaude mcp add skill-deck \  -- npx -y skill-deck mcp serve \  --skills ~/.codex/skills
bash
# guided：暴露 skill_guide，让模型先看分组和推荐入口claude mcp add skill-deck \  -- npx -y skill-deck mcp serve \  --skills ~/.codex/skills \  --skill-mode guided
bash
# compact：只暴露 list/search/info/read/resources，工具列表最短claude mcp add skill-deck \  -- npx -y skill-deck mcp serve \  --skills ~/.codex/skills \  --skill-mode compact

Codex / 其他 MCP host 使用同一组 stdio 参数；如需切换模式，在 args 中追加 --skill-mode 和对应值：

json
{  "mcpServers": {    "skill-deck": {      "command": "npx",      "args": [        "-y",        "skill-deck",        "mcp",        "serve",        "--skills",        "~/.codex/skills"      ]    }  }}

第二步，在 Claude / Codex 中确认 skill-deck 已连接。Claude Code 里输入：

text
/mcp

第三步，直接对 agent 说：

text
请使用 SkillDeck MCP 的 generate_skill_share 工具，为当前已加载的 ~/.codex/skills 生成一套可以分享给同行看的 Skill 工作流分享物。

需要指定输出目录时，再追加：

text
输出到 ~/skilldeck-shares/codex-skills-share，注意脱敏，并在完成后告诉我 index.html、cover.png、detail.png 和 manifest.json 的路径。

Share

分享不是让你手写配置，也不是把私有 Skill 扔到公开市场。你只需要和 agent 对话：

text
帮我把 ~/.skills 这套 Skills 做成一个可以发给同行看的分享页和两张分享图，注意脱敏。

agent 会通过 SkillDeck 的 MCP Share 接口生成：

text
share/  index.html  cover.png  detail.png

index.html 是完整介绍页，支持搜索和下钻到 Skill 详情；cover.png 适合传播，detail.png 适合给同行看技术细节。默认会移除本地路径、secret 和私有 URL，reference 只展示路径，不读取文件内容。

实际生成效果：

[SkillDeck share cover example] [SkillDeck share detail example]

全量能力

  • Skill Activation：compact / guided / active。
  • skill_guide 和 use_skill_*。
  • 搜索排序、分页、字段裁剪。
  • Catalog/resources。
  • watcher。
  • symlink 显式支持。
  • Share HTML + 双图。
  • docs-site 右上角搜索。
  • npm 包：skill-deck。
  • MCP Registry：server.json + npm/stdio 元数据。
  • MCP Streamable HTTP transport。

文档

更完整的在线文档见 skill.counterxing.top。

License

MIT

Source: README.md at commit a5c03a8

Tools

0
SourceWeft has not started this package in its sandbox yet, so its tools are not known.

Version history

1
  1. v1.0.0LatestSep 16, 2026

Score

50% of the applicable checks met

How complete this listing is, computed from the registry entry, the endpoint and public sources. Not a rating of the server and not a statement that it is safe or working today: reachability is a dated check, and no installs or reviews go into the score.

  • Reached in the last 30 days25 pts

    Keep the endpoint answering the MCP handshake; SourceWeft probes it every 30 days.

  • Auth declared10 pts

    Declare the credential headers in the registry entry, or keep the endpoint answering the probe so its auth can be classified.

  • Tools listed15 pts

    Answer tools/list without a credential, or ship a package that starts in a sandbox without one.

  • DescriptionRequired15 pts
  • Logo10 pts
  • Documentation15 pts
  • License10 pts