From content-vault
Transcribes a finished YouTube video and generates a description block with a short summary, fixed CTA, and chapter timestamps using ElevenLabs Scribe.
How this skill is triggered — by the user, by Claude, or both
Slash command
/content-vault:youtube-descriptionThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Turns a finished video cut into a copy-paste-ready YouTube description.
Turns a finished video cut into a copy-paste-ready YouTube description. Three parts, always in this order:
The transcription engine is ElevenLabs Scribe (word-level timestamps),
the same one /video-use uses. The output is a block you paste into
YouTube yourself (this skill does not touch the live video).
<YOUR_BRAND>, <YOUR_DOMAIN>, <YOUR_LINKEDIN_URL>. After that the
CTA is reproduced verbatim on every video, never reworded.ELEVENLABS_API_KEY in your env (or .env). Scribe needs it.
/video-use in this vault already uses the same key.<YOUR_BRAND_DOC> at your voice rules so the
description passes the same slop checks as your other copy.Trigger on:
/youtube-descriptionSibling skills (don't confuse):
/youtube-publisher/video-use/youtube-scriptA finished video. Accept any of:
.mp4 / .mov./video-use transcript on disk) —
reuse it to skip the API call and the wait. /video-use edits store
these under <videos_dir>/edit/transcripts/<stem>.json.Reuse a /video-use Scribe JSON if one exists (no API call, no wait —
Scribe on a 25-min video costs minutes and credits). Otherwise call
Scribe once and cache the result.
Either way, build a [M:SS] text transcript — one line per ~20s — and
note the total duration (no chapter may land at or past it). The
ElevenLabs Scribe endpoint returns word-level start/end times; group
words into ~20s buckets and prefix each line with its start timestamp.
Read the whole transcript. It is your only source for both the description and the chapters.
House format for a tutorial:
Learn how to set up a cold email system with Claude Code that can handle 100k+ emails / month. We'll be using Claude Code, <YOUR_BRAND>, and a set of other tools.
Pattern:
Learn how to <outcome> with <concrete spec>. The
outcome and the number are the hook. Lead with what the viewer can do.We'll be using <tool stack>. Lowercase the tool names,
end with "and other tools" / "and a set of custom skills" if the list
runs long.Rules: two sentences, no more. No em dashes (use . , : ( ·).
No banned words (leverage, seamless, powerful, … full list in
<YOUR_BRAND_DOC>). Concrete numerals over adjectives. It must read like
you typed it, not like a SaaS template.
From the transcript, pick chapter boundaries at real topic shifts and title each. Hard YouTube rules:
0:00. Always. Without it YouTube shows no chapters.M:SS, or H:MM:SS
past an hour).Exactly this structure. The fixed CTA block sits between the description and the timestamps and is reproduced verbatim — never reword, never drop a line, never change the handle:
<description · sentence 1>
<description · sentence 2>
Try <YOUR_BRAND> for free:
<YOUR_DOMAIN>
Connect with me on LinkedIn:
<YOUR_LINKEDIN_URL>
Timestamps:
0:00 <first chapter title>
<m:ss> <title>
<m:ss> <title>
...
Put the assembled block where your content lives so it's tracked and copy-pasteable:
## YouTube description
section in the video's .md file. Wrap the block in a fenced
```text code block so line breaks survive the paste into YouTube.
The section is idempotent — replace the existing one, don't append
a second.Report back where you wrote it and the chapter count. Show the assembled block in chat too.
0:00, none past the printed duration./youtube-publisher; here you explicitly
want the short description + timestamps. But "short" still means two
sentences.)/youtube-publisher's
finalize step (--only-descriptions) is the path./video-use JSON first and reuse it.npx claudepluginhub timscheuerai/content-vaultGenerates YouTube metadata (title, description, tags, chapters) from a transcript, topic, or description. Useful for content creators optimizing video SEO.
Writes a complete YouTube video description with search-optimized first lines, structured body, links, chapters, and call to action.
Fetches YouTube video transcripts, then summarizes, repurposes, or analyzes content. Use when a user pastes a YouTube URL for text-based output.