菜单
Fetch YouTube video transcripts. Use when the user asks to get a YouTube video transcript, subtitles, or captions, or wants to analyze/summarize a YouTube video.
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
Run and fix stylint checks for user-facing prose such as workshops, lessons, articles, public docs, and publishable agent-written text. Do not use for internal engineering documents, process docs, handoff docs, planning notes, or repo maintenance notes unless the user explicitly asks for prose linting or style polish.
Fetch and transcribe Google Recorder voice notes. Use when the user shares a recorder.google.com link and wants the original audio file, a transcript, or wants to act on a voice note.
Transcribe local audio files with the OpenAI Audio Transcriptions API without adding OpenAI dependencies to the target project.
Create and scaffold plugin directories for Codex with a required `.codex-plugin/plugin.json`, optional plugin folders/files, valid manifest defaults, and personal-marketplace entries by default. Use when Codex needs to create a new personal plugin, add optional plugin structure, generate or update marketplace entries for plugin ordering and availability metadata, or update an existing local plugin during development with the CLI-driven cachebuster and reinstall flow.
Initialize a new Python library with modern tooling, packaging, tests, and optional CLI support. Use when the user wants to scaffold a new Python package.
Add the standardized CI publish workflow (`.github/workflows/publish.yml` + `make release`) to an existing Python project so PyPI releases happen on tag push. Use when a project still publishes via a local script (`publish.py`, `hatch publish`, `twine upload`) or has no automated publish at all.
| name | fetch-youtube |
| description | Fetch YouTube video transcripts. Use when the user asks to get a YouTube video transcript, subtitles, or captions, or wants to analyze/summarize a YouTube video. |
| allowed-tools | Bash(uv run *) |
| argument-hint | ["video-url-or-id"] |
Fetch the transcript/subtitles of a YouTube video as timestamped text.
The script caches every transcript to ~/.cache/youtube_transcripts/<video-id>.txt and
is idempotent. So fetch once, then work from the file. Do not:
/tmp copy — the cache file is your copy,head/cat into context just to peek at it.Recommended: use --path to fetch+cache and get back only the file path, then use the
Read tool (with offset/limit for long videos) or grep on that file:
uv run --with youtube-transcript-api --with python-dotenv \
~/.claude/skills/fetch-youtube/youtube.py <video-id-or-url> --path
This prints the cache path to stdout (e.g. ~/.cache/youtube_transcripts/dQw4w9WgXcQ.txt)
and a fetched: <path> (<N> lines) status line to stderr — without sending the whole
transcript through bash. Then Read that file. For a short video you can read it whole; for a
long one, read in chunks or grep for the topic the user cares about.
Only stream the transcript to stdout when you genuinely need it inline and it's short:
uv run --with youtube-transcript-api --with python-dotenv \
~/.claude/skills/fetch-youtube/youtube.py <video-id-or-url>
The script accepts a video ID (dQw4w9WgXcQ), a full URL
(https://www.youtube.com/watch?v=dQw4w9WgXcQ), or a short URL (https://youtu.be/dQw4w9WgXcQ).
Timestamped subtitles, one line per cue:
0:00 Hello and welcome
0:05 Today we're going to talk about...
1:23:45 And that wraps up our discussion
Get all video IDs and titles from a playlist (tab-separated id<TAB>title):
python3 ~/.claude/skills/fetch-youtube/playlist.py <playlist-id-or-url>
Note: only the first page (~100 videos) is returned; longer playlists need pagination support added.
The script auto-loads Oxylabs proxy credentials from ~/.config/youtube/.env if present and
retries through the proxy on failure — no extra flags. The .env should contain:
OXYLABS_USER=...
OXYLABS_ENDPOINT=...
OXYLABS_PASSWORD=...
Proceed with what the user asked (summarize, extract key points, answer questions, make notes). If they only said "get the transcript," tell them where it is and ask what they'd like next.