菜单
Fetch transcripts from YouTube videos using yt-dlp. Supports any language with auto-generated or manual captions. Use for summarization, analysis, or translation tasks.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
| name | youtube-transcript |
| description | Fetch transcripts from YouTube videos using yt-dlp. Supports any language with auto-generated or manual captions. Use for summarization, analysis, or translation tasks. |
Fetch transcripts from YouTube videos using yt-dlp.
brew install yt-dlp jq
{baseDir}/transcript.sh <video-url-or-id> [language]
# English (default)
{baseDir}/transcript.sh "https://youtu.be/dQw4w9WgXcQ"
# Polish
{baseDir}/transcript.sh "https://youtu.be/ksWAT3uqlWA" pl
# German
{baseDir}/transcript.sh "https://www.youtube.com/watch?v=VIDEO_ID" de
# Just video ID works too
{baseDir}/transcript.sh dQw4w9WgXcQ
Timestamped transcript:
[0:00] [Music]
[0:18] We're no strangers to
[0:21] love. You know the rules...
[1:23] Never gonna give you up
Any language code: en, pl, de, es, ja, pt, fr, it, etc.
Use yt-dlp --list-subs <url> to see available languages for a video.
Analyze pi session transcripts to discover patterns that could become AGENTS.md rules, skills, or prompt templates. Mines your usage history for automation opportunities.
Build PR-focused context packs (PR description + git diff + full changed files + related files via Scribe), then budget-fit and token-check with o200k-base.
Review a pull request for bugs, style issues, and guideline compliance. Local review - does not post to GitHub.
Build LLM-ready code dumps with optional docs, then count tokens with tokencount (o200k-base) against a context budget (for example 272k for GPT-5).
Compress images to a target file size using macOS sips. No dependencies required.
Multi-model code review. Runs code-review skill with 3 models in parallel, then synthesizes findings.