YouTube Podcast Skill

Convert a YouTube podcast into structured local knowledge: a cleaned timestamped transcript, a summary + takeaways, a mindmap canvas, and an SEO-optimized blog post.

Output Goal

When a user provides a YouTube URL, produce exactly these files under podcasts/<video-title>/ (current working directory):

• •podcasts/<video-title>/transcript.md
• •podcasts/<video-title>/summary.md
• •podcasts/<video-title>/mindmap.canvas
• •podcasts/<video-title>/blog-post.md

Workflow (End-to-End)

1. •

   Fetch Video Title

   • •Use the YouTube oEmbed API: https://www.youtube.com/oembed?url=<YOUTUBE_URL>&format=json.
   • •Fetch the JSON content and extract the title field.
   • •Sanitize for folder name: replace <>:"/\\|?* with -, collapse whitespace, trim.
2. •

   Create Output Folder

   • •Create podcasts/<video-title>/ under the current working directory.
3. •

   Extract Transcript (Robust Method)

   • •Navigate to https://www.youtube-transcript.io/ using agent-browser.
   • •Input the YouTube URL and trigger extraction.
   • •Use the Chunked Redirection Strategy to save the full text to podcasts/<video-title>/raw.txt.
   • •Critical: Follow the specific loop-based extraction in references/transcript-extraction.md to avoid truncation.
4. •

   Clean Transcript (Markdown)

   • •Run scripts/clean_transcript.py <raw.txt> -o <transcript.md>.
   • •Enforce one line per timestamp: timestamp + sentence only.
5. •

   Generate Summary + Takeaways (Markdown)

   • •Read transcript.md and write summary.md using this fixed structure:
     • •# Summary (1–3 paragraphs)
     • •# Takeaways (5–10 bullets)
6. •

   Generate Mindmap (JSON Canvas)

   • •Locally invoke the json-canvas skill to generate the file.
   • •Use the full content of transcript.md as context.
   • •Save as mindmap.canvas.
7. •

   Generate SEO Blog Post (Markdown)

   • •Read transcript.md and write blog-post.md following SEO best practices (compelling H1, meta description, H2/H3 structure, impactful quotes, and CTA).

Example Commands

# 1. Fetch Title
curl -s "https://www.youtube.com/oembed?url=<URL>&format=json"

# 2. Extract Transcript (Robust Loop)
agent-browser open "https://www.youtube-transcript.io/"
# (Find refs, fill, click, wait...)
# Measure length
agent-browser eval "document.body.innerText.length"
# Run chunked loop (see references/transcript-extraction.md for details)
agent-browser eval "document.body.innerText.substring(0, 15000)" > "raw.txt"
agent-browser eval "document.body.innerText.substring(15000, 30000)" >> "raw.txt"
# ...repeat until end

# 3. Clean
python "scripts/clean_transcript.py" "raw.txt" -o "transcript.md"