youtube-skills README

项目的 README,来自 github.com/artemchuikin/youtube-skills。合并为一个 markdown 文件:README.md。

YouTube Skills for AI Agents 🎬

Drop-in skills that let a coding or chat agent pull YouTube transcripts, run searches, walk channels and export whole playlists.

An Agent Skill is a folder of instructions an agent loads when the task matches. The 12 skills in this repo cover YouTube end to end. Each one names its endpoint, shows the exact request, lists every parameter with defaults and limits, and maps every error code to a recovery, so the agent makes the right call on the first try instead of guessing. No yt-dlp, no headless browser, no Google Cloud project. One HTTPS request per answer.

Works with Claude Code, Cursor, Antigravity, Cline, Codex, OpenClaw, Hermes Agent, and anything else that reads the Agent Skills format.

Made by TranscriptOut · API docs · MCP server


⚡ Quick start

npx skills add artemchuikin/youtube-skills --skill youtube-full

Then ask the agent something it previously had to refuse.

What does this talk actually claim? https://www.youtube.com/watch?v=dQw4w9WgXcQ

The first run also takes care of the API key, without leaving the chat. The agent asks for your email, a 6-digit code arrives in your inbox, you read it back, and a fresh sk_ key lands in your shell or agent config. The account behind it is free, starts with 100 credits and never asks for a card.

Other install paths

Hermes Agent

hermes skills install skills-sh/artemchuikin/youtube-skills/skills/youtube-full

All 12 skills at once

npx skills add artemchuikin/youtube-skills

Manual (git clone)

git clone https://github.com/artemchuikin/youtube-skills.git
cp -r youtube-skills/skills/youtube-full ~/.claude/skills/

OpenClaw (ClawdBot/Moltbot). Copy a folder from clawhub/ into your skills directory. The clawhub/ tree carries the same 12 skills with bash-style examples.

No terminal at all. Paste this into the agent and let it do the work.

Install the youtube skills from this GitHub repo: https://github.com/artemchuikin/youtube-skills
I want to get YouTube transcripts, search YouTube, and browse channel videos from my agent.
Set everything up for me.

Installing skills and creating the API key are both things these skills teach the agent to do.


💬 What you can ask

Every row is something you can say to the agent word for word. The skill turns it into the right API call.

Ask for For example
A transcript "What does this video actually say? [URL]"
A summary or quote "Pull the three strongest quotes from this talk"
Subtitles as files "Export this video's subtitles as SRT"
A YouTube search "Find recent talks about Rust async and pick the best three"
A channel's news "Anything new on @kurzgesagt this week?"
A channel's catalogue "List everything @3blue1brown has ever uploaded"
Search inside a channel "Find every Fireship video that mentions Postgres"
A whole playlist "Turn this lecture course into study notes"
A bulk export "Queue transcripts for all 800 videos on this channel"
RAG-ready chunks "Fetch this playlist's transcripts in ~1000-character segments for my index"

🧰 Which skill to install

youtube-full is the right answer for most people. One skill, the whole surface. The narrower ones exist because a skill is context the agent has to carry, and a workflow that only ever needs transcripts should not pay for playlist documentation.

Skill What it does Install
youtube-full Everything. Transcripts, search, channels, playlists, bulk jobs --skill youtube-full
transcript Video transcripts with timestamps --skill transcript
youtube-search Search YouTube for videos and channels --skill youtube-search
youtube-channels Uploads, latest videos, search within a channel --skill youtube-channels
youtube-playlist Every video of a playlist --skill youtube-playlist
# one skill
npx skills add artemchuikin/youtube-skills --skill transcript

# several at once
npx skills add artemchuikin/youtube-skills --skill transcript --skill youtube-search
All 12 skills

Beyond the core five, the rest are vocabulary variants of the same capabilities. People say "captions", "subtitles", "video to text" or "youtube api" for overlapping things, and an agent picks a skill by matching your words against skill descriptions, so each phrasing has a skill that answers to it.

Skill Description
youtube-full Complete toolkit. Transcripts + search + channels + playlists + bulk jobs
transcript Extract the transcript of any YouTube video, with timestamps
captions Closed captions / CC, including SRT and VTT files
subtitles Subtitles from YouTube videos, including SRT and VTT files
youtube-search Search YouTube for videos and channels
youtube-channels Channel uploads, latest videos, in-channel search
youtube-playlist All videos of a YouTube playlist
youtube-data Video and channel data without Google's API quotas
youtube-api YouTube API access for agents, no Google Cloud project
transcriptout The full API surface in one skill
video-transcript Convert YouTube videos to text
yt Quick lookups. Transcript, search, channel latest

💳 What calls cost

The unit is a credit, and one credit is one answered request.

Request Cost
transcript of one video (any of 5 formats) 1
video metadata + available caption languages 1
YouTube search 1 per page
channel latest (~15 newest videos) 1
channel videos 1 per page (100 videos, or 500 IDs)
search within a channel 1 per page
playlist videos 1 per page (100 videos, or 500 IDs)
search within a playlist 1
bulk job submit 1 per video
bulk job progress, results, cancel free

A call that dies before reaching YouTube (bad parameter, rate limit, service busy) refunds itself automatically. A confirmed "this video has no captions" costs the credit, because the lookup really happened. The rate limit is 200 requests per minute per key on every plan.

Plan Price Credits
Free $0 100 on signup, one-time
Starter $4.49/month 1,000/month
Starter Annual $45.29/year (~$3.77/mo) 1,000/month
Scale slider up to $198.99/mo up to 100,000/month

Subscription volume is a slider from 1,000 to 100,000 credits per month in steps of 1,000, and the per-1,000 rate falls as the volume grows. Full pricing →


📦 Built for bulk work

Bulk export is one job, not a loop. The agent collects video IDs cheaply (ids_only=true returns up to 500 per page), submits them all in a single POST /v1/transcripts (up to 4,000 videos), and reads results page by page while the job is still running. Duplicates are dropped before billing, an Idempotency-Key makes retries free, the job paces itself inside the rate limit instead of tripping 429s, and any video the service fails to deliver is refunded on its own. A full channel archive that used to be an evening of scripting is three prompts.

🧠 Chunks sized for retrieval

Transcripts arrive pre-chunked if you ask. The segment parameter (20 to 5,000 characters) controls how the text is cut, and 500 to 1,500 gives retrieval-sized pieces with real context. Cut points come from per-word timestamps in YouTube's raw srv3 data, so each chunk's start is the moment the words are actually spoken rather than an interpolation. segment=1000&format=json goes straight into an embedding pipeline with no re-chunking step.


🔑 Setting up the key by hand

The in-chat flow above is optional. The manual way takes two minutes.

  1. Sign in at transcriptout.com with Google or a magic link. The first sign-in is the signup and grants the free credits.
  2. Create an API key in the dashboard. It starts with sk_ and is shown once.
  3. Store it where the agent can read it, without pasting it into the chat. If your runtime asks for TRANSCRIPTOUT_API_KEY itself (Hermes, OpenClaw), enter it there. Otherwise run this in your own terminal: it reads the key without echoing it into a file only you can read.
mkdir -p ~/.config/transcriptout && chmod 700 ~/.config/transcriptout && (umask 077; printf 'TranscriptOut key: '; stty -echo; IFS= read -r k; stty echo; echo; printf '%s' "$k" > ~/.config/transcriptout/key)
  1. Load it in every shell. Add this line to your shell profile. It holds a path, not the key:
export TRANSCRIPTOUT_API_KEY="$(cat ~/.config/transcriptout/key 2>/dev/null)"
Where does the key get saved?
Location File
OpenClaw/Moltbot the runtime's own config, via its prompt for TRANSCRIPTOUT_API_KEY
Hermes Agent Hermes secret store (TRANSCRIPTOUT_API_KEY declared via required_environment_variables in skill frontmatter)
Everything else ~/.config/transcriptout/key (folder mode 700, file mode 600), loaded by one line in the shell profile

The key itself never goes into a shell profile, a project directory or the chat, and the skills never print it. If you pasted it into a chat anyway, create a fresh key in the dashboard and delete the old one.


❓ FAQ

Can I bulk download YouTube transcripts?

Yes, and there is a purpose-built endpoint for it. Install youtube-full and ask for transcripts of a playlist or channel. The agent gathers the IDs and submits one batch job for up to 4,000 videos, then reads results as they land. Same price per video as single calls.

Do I need a Google YouTube Data API key?

No. These skills replace the Data API for transcripts, search, and channel and playlist listings, without a Google Cloud project, quota units or OAuth screens. Google's API also does not return transcripts, which tends to be the whole point.

Do I need an API key at all?

Yes, but getting one is part of the first run. The agent registers your email, you read a 6-digit code off your inbox, and the key is stored for you. 100 free credits, no card.

Which skill should I install?

youtube-full, unless you know you want less. The narrow ones are listed under Which skill to install.

What API is underneath?

TranscriptOut, a hosted YouTube data API. The full reference lives at transcriptout.com/docs, and the same backend is available as a remote MCP server if your client speaks MCP natively. For code outside an agent there are client libraries, pip install transcriptout and npm i transcriptout, drop-in replacements for youtube-transcript-api and youtube-transcript.

Live streams?

Transcripts appear after a stream ends and captions are processed, not while it runs.


🧯 Troubleshooting

Every error response carries a machine-readable code and a request_id, and agents tend to relay them word for word, so this table is keyed by what you will actually see.

Symptom Likely cause Fix
401 Unauthorized Key missing, mistyped, or not loaded into the environment Confirm TRANSCRIPTOUT_API_KEY is set in the shell the agent runs in and starts with sk_
402 Payment Required The account is out of credits Check the balance at transcriptout.com/billing
404 Not Found No captions on the requested language or track, or a wrong ID A definitive answer, retrying won't change it. Try another lang, or kind=auto
410 Gone The video was removed Nothing to fetch
451 Age-restricted or members-only video Nothing to fetch without an account, which the API deliberately does not use
422 / 400 A malformed parameter Channels accept @handle, a name, a URL or a UC ID. Playlists take a PL... ID or URL. The credit refunds itself
429 Too Many Requests Rate limit reached (200 req/min per key) Wait out the Retry-After header. The credit refunds itself
503 Temporary capacity pressure on cold fetches Retry after Retry-After. The credit refunds itself

And three situations that are not errors at all.


🗂️ Repository layout

skills/     the 12 skills with HTTP-style examples (Claude Code, Cursor, Codex, Hermes, ...)
clawhub/    the same 12 with bash examples, for OpenClaw
scripts/    generate.py, the single source every skill file is emitted from
assets/     icons

The skill files are generated. Edit scripts/generate.py and re-run it rather than touching skills/ or clawhub/ by hand. The endpoints and error tables stay consistent across all 12 because they come from one section library.

🤝 Contributing

PRs welcome. See CONTRIBUTING.md.

📄 License

MIT. These skills call the TranscriptOut API, an independent product that is not affiliated with or endorsed by YouTube or Google.

youtube-skills (artemchuikin/youtube-skills)

正在搜索数据库…

这是什么?这是一组 Agent Skills(智能体技能)——也就是指令文件夹,可以让 Claude Code、Cursor、Codex 等编程或聊天智能体通过 TranscriptOut API 获取 YouTube 字幕稿、搜索视频、读取频道和播放列表,并执行批量导出。
这是我需要导入的库,还是别的东西?这是一组 Agent Skills(智能体技能)——当任务与之匹配时,智能体会自动加载的指令文件夹;它不是需要你导入的库,也不是需要你启动的应用。
我该如何在智能体中安装这些技能?在智能体中运行类似 `--skill youtube-full` 的命令安装整个套件(或者安装范围更小的技能,例如 `--skill transcript`);如果你使用 OpenClaw、ClawdBot 或 Moltbot,则将 `clawhub/` 中对应的文件夹复制到你的技能目录,或者手动克隆仓库。
使用它需要谷歌 API 密钥或其他东西吗?不需要谷歌云项目;你只需要一个 TranscriptOut API 密钥(以 sk 开头)。首次使用时,智能体可以通过一次性邮件验证码为你申请,附赠 100 点免费额度,无需绑定银行卡。
有没有某些视频根本无法获取的情况?直播的字幕稿要等直播结束、字幕处理完成后才会出现;年龄限制或仅会员可看的视频会返回 451 错误,因为 API 刻意不使用账号来绕过这些限制。
哪些编程智能体真正能与它配合使用?它适用于 Claude Code、Cursor、Antigravity、Cline、Codex、OpenClaw、Hermes Agent,以及其他能读取 Agent Skills 格式的智能体。
youtube-full 与范围更小的技能有什么区别?youtube-full 包含全部功能(字幕稿、搜索、频道、播放列表、批量任务);而 transcript 或 youtube-search 这类技能只包含某一项功能的指令,这样智能体就不会加载用不到的上下文。
我能批量导出整个频道的字幕稿吗?可以。智能体每页最多收集 500 个视频 ID,通过 POST /v1/transcripts 一次性提交最多 4,000 个视频的批量任务,并在任务运行时逐页读取结果;单个视频的价格与单次调用相同。
我还需要官方的 YouTube Data API 吗?不需要。这些技能可以完全替代它,用于获取字幕稿、搜索以及列出频道或播放列表,无需谷歌云项目、配额单位或 OAuth 授权页面;而且与官方 API 不同,它们确实能返回字幕稿。
我能获取为 RAG 流程预先分块的字幕稿吗?可以。segment 参数(20 至 5,000 个字符)会在真实的逐词时间戳处切分字幕稿,因此像 segment=1000&format=json 这样的调用可以直接输入嵌入索引,无需额外的重新分块步骤。
大小
0.1 MB
星标
275
更新
提交Nov 05
时长
2个月
提交 · 90 天