Gustavo Paiva

Youtube

AI 智能体v1.0.1
作者:Gustavo Paiva更新于 2026-05-17共 2 个版本 · 最新版本发布于 2026-01-28youtube@clawhub_grpaiva/youtube
工具调用#Transcription

据 AlphaMoat 统计,截至 2026-10,Youtube 累计下载 12,852 次、获 6 个星标,按累计下载量在本次参与排名的 10,914 个 AI 技能中排名第 2,171 位,在「AI 智能体」分类(1,632 个)中排名第 294 位。

列表默认按累计下载量排序,也可按星标排序;详情页排名按累计下载量计算。

累计下载
1.3万
星标
6
全站下载量排名
#2,171
全站前 20% · 「AI 智能体」前 19%
平台累计下载星标评论
ClawHub ↗9,7166-
SkillHub ↗1.3万6—

各平台的计数口径不同,统计时间不一致。

这个技能是做什么的

功能:视频详情(标题、描述、数据统计、发布日期)、带时间戳的转录文本、频道信息与近期视频、站内搜索、播放列表信息。安装:npm install -g zubeid-youtube-mcp-server(主方法)与 pip install yt-dlp(兜底),API key 从 Google Cloud Console 创建(启用 YouTube Data API v3,建议把 key 限制到该 API),经 ~/.clawdbot/clawdbot.json 配置或 export YOUTUBE_API_KEY 传入。调用走 mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js",支持 search_videos、channels_info、channels_listVideos、videos_details、transcripts_getTranscript 等方法;MCP 转录失败(空或不可用)时用 yt-dlp --write-auto-sub 抓取 vtt。API 配额:默认每日 10,000 units,search 100 units/次、视频详情 1 unit/次、transcript 0 units(独立机制)——建议转录放心用、search 节制用。安全:key 只用于官方 YouTube Data API、无第三方服务器、全部网络调用指向 googleapis.com;key 保存在 Clawdbot 配置而非代码中,不要提交到 git。注意转录可能是自动生成,引用前核对准确性。

资料发布于 2026-10-27

可以用来做什么

  • 搜索 YouTube 视频并按主题/频道筛选结果
  • 获取视频详情、频道近期视频列表与播放列表信息
  • 抓取视频字幕转录文本用于研究或内容分析(MCP 失败时走 yt-dlp 兜底)

什么时候用

  • 开发者、研究者与内容分析者用它检索 YouTube、获取元数据与转录
  • 查找播客最新一期、为研究获取转录文本等常见工作流

示例提示词

以下是根据技能文档生成的示例,不代表实际运行结果。

搜一下关于 Laravel AI 效率工具的最新视频
帮我拿到这期 Lex Fridman 播客的完整字幕
这个频道最近发了哪些视频?

月度数据趋势

累计下载1.3万
6,6909,9991.3万2026-05 · 累计下载 7,1462026-06 · 累计下载 7,5912026-07 · 累计下载 8,5502026-08 · 累计下载 9,1372026-09 · 累计下载 9,4292026-10 · 累计下载 1.3万26-0526-0926-10
星标6
5672026-05 · 星标 62026-06 · 星标 62026-07 · 星标 62026-08 · 星标 62026-09 · 星标 62026-10 · 星标 626-0526-0926-10

2026-09 起统计方法调整:同名技能按作者分别统计,并覆盖全部技能。图中虚线为调整前的月份,与之后的数据不直接比较,也不计算跨越调整的增长。 该技能有同名的其他版本,调整前的数据归属当前下载量最大的这一个。

月份累计下载环比星标
2026-101.3万—6
2026-099,429—6
2026-089,137+6.9%6
2026-078,550+12.6%6
2026-067,591+6.2%6
2026-057,146—6

SKILL.md 原文

以下是作者发布的原始文件,按纯文本显示,本站未做修改或审核;其中的链接不可点击,命令请核实后再运行。抓取于 2026-10-23。

7.3 KB · ClawHub v1.0.1
---
name: youtube
description: Search YouTube videos, get channel info, fetch video details and transcripts using YouTube Data API v3 via MCP server or yt-dlp fallback.
metadata: {"clawdbot":{"emoji":"📹","requires":{"bins":["yt-dlp"],"npm":["zubeid-youtube-mcp-server"]},"primaryEnv":"YOUTUBE_API_KEY"}}
---

# YouTube Research & Transcription

Search YouTube, get video/channel info, and fetch transcripts using YouTube Data API v3.
展开全部(共 297 行)
## Features

- 📹 Video details (title, description, stats, publish date)
- 📝 Transcripts with timestamps
- 📺 Channel info and recent videos
- 🔍 Search within YouTube
- 🎬 Playlist info

## Setup

### 1. Install dependencies

**MCP Server (primary method):**
```bash
npm install -g zubeid-youtube-mcp-server
```

**Fallback tool (if MCP fails):**
```bash
# yt-dlp for transcript extraction
pip install yt-dlp
```

### 2. Get YouTube API Key

1. Go to [Google Cloud Console](https://console.cloud.google.com)
2. Create/select a project (e.g., "YouTube Research")
3. Enable the API:
   - Menu → "APIs & Services" → "Library"
   - Search: "YouTube Data API v3"
   - Click "Enable"
4. Create credentials:
   - "APIs & Services" → "Credentials"
   - "Create Credentials" → "API Key"
   - Copy the key
5. Optional - Restrict:
   - Click the created key
   - "API restrictions" → Select only "YouTube Data API v3"
   - Save

### 3. Configure API Key

**Option A: Clawdbot config** (recommended)
Add to `~/.clawdbot/clawdbot.json`:
```json
{
  "skills": {
    "entries": {
      "youtube": {
        "apiKey": "AIzaSy..."
      }
    }
  }
}
```

**Option B: Environment variable**
```bash
export YOUTUBE_API_KEY="AIzaSy..."
```

### 4. Setup MCP Server

The skill will use `mcporter` to call the YouTube MCP server:

```bash
# Build from source (if installed package has issues)
cd /tmp
git clone https://github.com/ZubeidHendricks/youtube-mcp-server
cd youtube-mcp-server
npm install
npm run build
```

## Usage

### Search Videos

```bash
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  search_videos query="ClawdBot AI" maxResults:5
```

Returns video IDs, titles, descriptions, channel info.

### Get Channel Info

```bash
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  channels_info channelId="UCSHZKyawb77ixDdsGog4iWA"
```

### List Recent Videos from Channel

```bash
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  channels_listVideos channelId="UCSHZKyawb77ixDdsGog4iWA" maxResults:5
```

### Get Video Details

```bash
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  videos_details videoId="Z-FRe5AKmCU"
```

### Get Transcript (Primary)

```bash
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  transcripts_getTranscript videoId="Z-FRe5AKmCU"
```

### Get Transcript (Fallback with yt-dlp)

If MCP transcript fails (empty or unavailable), use `yt-dlp`:

```bash
yt-dlp --skip-download --write-auto-sub --sub-lang en --sub-format vtt \
  --output "/tmp/%(id)s.%(ext)s" \
  "https://youtube.com/watch?v=Z-FRe5AKmCU"
```

Then read the `.vtt` file from `/tmp/`.

**Or get transcript directly:**
```bash
yt-dlp --skip-download --write-auto-sub --sub-lang en --print "%(subtitles)s" \
  "https://youtube.com/watch?v=VIDEO_ID" 2>&1 | grep -A1000 "WEBVTT"
```

## Common Workflows

### 1. Find Latest Episode from a Podcast

**Example: Lex Fridman Podcast**

```bash
# Get channel ID (Lex Fridman: UCSHZKyawb77ixDdsGog4iWA)
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  channels_listVideos channelId="UCSHZKyawb77ixDdsGog4iWA" maxResults:1
```

Returns most recent video with title, ID, publish date.

### 2. Get Transcript for Research

```bash
# Step 1: Get video ID from search or channel listing
# Step 2: Try MCP transcript first
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  transcripts_getTranscript videoId="VIDEO_ID"

# Step 3: If empty, fallback to yt-dlp
yt-dlp --skip-download --write-auto-sub --sub-lang en \
  --output "/tmp/%(id)s.%(ext)s" \
  "https://youtube.com/watch?v=VIDEO_ID"

cat /tmp/VIDEO_ID.en.vtt
```

### 3. Search for Topics

```bash
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  search_videos query="Laravel AI productivity 2025" maxResults:10
```

Filter results for relevant channels or dates.

## Channel IDs Reference

Keep frequently used channels here for quick access:

- **Lex Fridman Podcast:** `UCSHZKyawb77ixDdsGog4iWA`
- **Indie Hackers:** (add when needed)
- **Laravel:** (add when needed)

To find a channel ID:
1. Go to channel page
2. View page source
3. Search for `"channelId":` or `"externalId"`

Or use search and extract from results.

## API Quota Limits

YouTube Data API v3 has daily quotas:
- Default: 10,000 units/day
- Search: 100 units per call
- Video details: 1 unit per call
- Transcript: 0 units (uses separate mechanism)

**Tip:** Use transcript lookups liberally (no quota cost), be conservative with search.

## Troubleshooting

### MCP Server Not Working

**Symptom:** `Connection closed` or `YOUTUBE_API_KEY environment variable is required`

**Fix:** Build from source:
```bash
cd /tmp
git clone https://github.com/ZubeidHendricks/youtube-mcp-server
cd youtube-mcp-server
npm install
npm run build

# Test
YOUTUBE_API_KEY="your_key" node dist/cli.js
```

### Empty Transcripts

**Symptom:** Transcript returned but content is empty

**Cause:** Video may not have captions, or MCP can't access them

**Fix:** Use yt-dlp fallback (see above)

### yt-dlp Not Found

```bash
pip install --user yt-dlp
# or
pipx install yt-dlp
```

## Security Note

The YouTube API key is safe to use with this MCP server:
- ✅ Key only used to authenticate with official YouTube Data API
- ✅ No third-party servers involved
- ✅ All network calls go to `googleapis.com`
- ✅ Code reviewed (no data exfiltration)

However:
- 🔒 Keep the key in Clawdbot config (not in code/scripts)
- 🔒 Restrict API key to YouTube Data API v3 only (in Google Cloud Console)
- 🔒 Don't commit the key to git repositories

## Examples

### Research Podcast for LinkedIn Post Ideas

```bash
# 1. Find latest Lex Fridman episode
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  channels_listVideos channelId="UCSHZKyawb77ixDdsGog4iWA" maxResults:1

# 2. Get video details
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  videos_details videoId="Z-FRe5AKmCU"

# 3. Get transcript
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  transcripts_getTranscript videoId="Z-FRe5AKmCU"

# If transcript empty, use yt-dlp
yt-dlp --skip-download --write-auto-sub --sub-lang en \
  --output "/tmp/%(id)s.%(ext)s" \
  "https://youtube.com/watch?v=Z-FRe5AKmCU"

# 4. Analyze transcript for interesting topics
# (read /tmp/Z-FRe5AKmCU.en.vtt and extract key themes)
```

### Find Videos About a Trending Topic

```bash
# Search for recent videos
mcporter call --stdio "node /tmp/youtube-mcp-server/dist/cli.js" \
  search_videos query="ClawdBot security concerns" maxResults:10

# Pick relevant ones, get transcripts
# Analyze sentiment and technical claims
```

## Notes

- MCP server path: `/tmp/youtube-mcp-server/dist/cli.js`
- Always pass API key via environment: `YOUTUBE_API_KEY="key" node ...`
- Or set globally in shell/Clawdbot config
- Transcripts may be auto-generated (check accuracy for quotes)
- yt-dlp can also download audio if you need it (`--extract-audio --audio-format mp3`)

「AI 智能体」分类下的热门技能

数据月份:2026-10 · 下载/星标数聚合自公开技能注册库(ClawHub、SkillHub),按月更新。口径说明见数据方法。