skills MCP — Anwendungsfälle, Installieren & Live-Demo

Name: skills MCP Server
Author: aahl

Warum nutzen

Hauptfunktionen

Home Assistant 智能家居控制
TTS: 微软 Edge 与智谱 GLM-TTS
DuckDuckGo 搜索与 DeepWiki 文档检索
加密货币行情与天气预报
Lark/飞书工作流
影视搜索、商品比价

Live-Demo

In der Praxis

skills-skill-3.replay ▶ bereit

0/0

Installieren

Wählen Sie Ihren Client

~/Library/Application Support/Claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "skills-skill-3": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/aahl/skills",
        "~/.claude/skills/skills"
      ],
      "_inferred": true
    }
  }
}

Öffne Claude Desktop → Settings → Developer → Edit Config. Nach dem Speichern neu starten.

~/.cursor/mcp.json · .cursor/mcp.json

{
  "mcpServers": {
    "skills-skill-3": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/aahl/skills",
        "~/.claude/skills/skills"
      ],
      "_inferred": true
    }
  }
}

Cursor nutzt das gleiche mcpServers-Schema wie Claude Desktop. Projektkonfiguration schlägt die globale.

VS Code → Cline → MCP Servers → Edit

{
  "mcpServers": {
    "skills-skill-3": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/aahl/skills",
        "~/.claude/skills/skills"
      ],
      "_inferred": true
    }
  }
}

Klicken Sie auf das MCP-Servers-Symbol in der Cline-Seitenleiste, dann "Edit Configuration".

~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "skills-skill-3": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/aahl/skills",
        "~/.claude/skills/skills"
      ],
      "_inferred": true
    }
  }
}

Gleiche Struktur wie Claude Desktop. Windsurf neu starten zum Übernehmen.

~/.continue/config.json

{
  "mcpServers": [
    {
      "name": "skills-skill-3",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/aahl/skills",
        "~/.claude/skills/skills"
      ]
    }
  ]
}

Continue nutzt ein Array von Serverobjekten statt einer Map.

~/.config/zed/settings.json

{
  "context_servers": {
    "skills-skill-3": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/aahl/skills",
          "~/.claude/skills/skills"
        ]
      }
    }
  }
}

In context_servers hinzufügen. Zed lädt beim Speichern neu.

claude mcp add skills-skill-3 -- git clone https://github.com/aahl/skills ~/.claude/skills/skills

Einzeiler. Prüfen mit claude mcp list. Entfernen mit claude mcp remove.

Anwendungsfälle

Praxisnahe Nutzung: skills

How to automate a smart-home routine via Home Assistant

👤 Home Assistant users wanting voice-free automation via Claude ⏱ ~20 min intermediate

Wann einsetzen: You want to chain actions (lights + AC + music) with conditional logic.

Voraussetzungen

Skill installed — git clone https://github.com/aahl/skills ~/.claude/skills/aahl-skills
Home Assistant URL + long-lived token — HA user profile → create token; export HA_URL, HA_TOKEN

Ablauf

List entities

列出所有与 'living_room' 相关的 Home Assistant entities。✓ Kopiert

→ Entity list with current states
Define routine

我晚上 7 点到家。灯调至 60%、空调打开到 26°C、播放轻音乐。✓ Kopiert

→ Sequenced service calls with expected outcome
Execute

执行这个 routine，按步骤 echo 服务调用后再执行。✓ Kopiert

→ Per-step confirmation + resulting state

Ergebnis: A working chained automation without hand-writing HA YAML.

Fallstricke

Wrong entity selected (similar names) — Require the skill to confirm entity_id before executing service call

Generate voice notes from markdown

👤 Content creators producing audio versions ⏱ ~10 min beginner

Wann einsetzen: You have a blog post and want a quick audio version for WeChat or podcast RSS.

Ablauf

Pick voice

用 Edge TTS 的 zh-CN-XiaoxiaoNeural 把 post.md 合成 mp3。✓ Kopiert

→ Audio file produced
Upgrade for emotion

改用 GLM-TTS，声音要更自然有起伏。✓ Kopiert

→ Alternate audio with different voice

Ergebnis: An audio version of your post ready to publish.

Post a daily digest to Lark/Feishu

👤 Team leads running Lark-based teams ⏱ ~25 min intermediate

Wann einsetzen: Each morning you want an automated team digest.

Voraussetzungen

Lark app with message-send permission — feishu.cn open platform → app → bot → add to group

Ablauf

Compose

今天的要闻：天气、加密货币BTC/ETH价格、团队昨日 GitHub PR 摘要。✓ Kopiert

→ Composed message
Send

发送到 '开发组' 群。✓ Kopiert

→ Success ack from Lark API

Ergebnis: A daily digest landing in Lark without manual copy-paste.

Compare prices before buying

👤 Shoppers on Chinese e-commerce platforms ⏱ ~5 min beginner

Wann einsetzen: Before clicking buy on a product across 淘宝/京东/拼多多.

Ablauf

Query

比价 'iPhone 15 Pro 256GB' 在主要平台，附链接。✓ Kopiert

→ Platform-by-platform price + URL

Ergebnis: A price comparison with direct links.

Fallstricke

Cached / stale prices — Skill should fetch live; verify with a click-through

Kombinationen

Mit anderen MCPs für 10-fache Wirkung

skills-skill-3 + github

Pull GitHub PR data then use AAHL skills to post a digest to Lark/Feishu

Use GitHub MCP to list this week's merged PRs in our org, then use AAHL's Lark skill to post a summary to the dev group.✓ Kopiert

skills-skill-3 + filesystem

Read a local markdown file and convert it to audio with AAHL's TTS skill

Read ~/blog/latest-post.md, then use AAHL's Edge TTS to generate a Chinese audio version.✓ Kopiert

Werkzeuge

Was dieses MCP bereitstellt

Werkzeug	Eingaben	Wann aufrufen	Kosten
Home Assistant control	-	Smart-home actions	HA API (local)
TTS (Edge / GLM)	text, voice	Voice content	Edge free / GLM paid
DuckDuckGo search	query	Web lookups	0
DeepWiki docs	query	Reference lookups	0
Crypto + weather	symbol or location	Daily digests	0
Lark/Feishu messaging	group, text	Team ops	Lark API free
Movie + price search	query	Daily life	0

Kosten & Limits

Was der Betrieb kostet

API-Kontingent: Free for most; GLM-TTS and some search APIs bring-your-own-key
Tokens pro Aufruf: Low per tool call
Kosten in €: Mostly free
Tipp: Use Edge TTS (free) unless you need GLM's emotion.

Sicherheit

Rechte, Secrets, Reichweite

Credential-Speicherung: Various: HA_TOKEN, GLM API key, Lark app credentials — all in env

Datenabfluss: Home Assistant (local LAN), Edge TTS (Microsoft), GLM (Zhipu), Lark (Feishu), DuckDuckGo

Home Assistant long-lived tokens grant full control of your house. Scope to a user with limited entities if possible.
Lark app permissions should be scoped to the minimum channels needed.

Fehlerbehebung

Häufige Fehler und Lösungen

Home Assistant 401

Token expired or URL wrong (check http vs https, port 8123)

Prüfen: curl -H "Authorization: Bearer $HA_TOKEN" $HA_URL/api/

Lark send fails

Bot not added to target group, or tenant_access_token expired. Re-issue.

Edge TTS silent on Linux

Install ffmpeg; Edge TTS depends on it for mp3 output

Prüfen: ffmpeg -version

Alternativen

skills vs. andere

Alternative	Wann stattdessen	Kompromiss
Home Assistant MCP	You want MCP semantics rather than skill prompts	Different integration shape

Mehr

Ressourcen

📖 Offizielle README auf GitHub lesen

🐙 Offene Issues ansehen

🔍 Alle 400+ MCP-Server und Skills durchsuchen