YouTube Channel Transcripts

A multi-channel archive of YouTube transcripts. Each channel lives in its own folder under channels/ with timestamped transcripts of (typically) the first 20 videos.

Repository Structure

channels/
  <channel-slug>/
    README.md             # channel description + index of videos
    transcripts/
      01-<video-title-slug>.md
      02-<video-title-slug>.md
      ...
README.md                 # this file

Channels

Channel	Folder	Videos
AI Engineer Europe	channels/ai-engineer-europe	24 (with analysis notes)
Lenny's Podcast	channels/lennys-podcast	19 (with notes)
Greg Isenberg	channels/greg-isenberg	20
My First Million	channels/my-first-million	80 (with notes)

Agent Instructions — Adding a New Channel

When asked to "transcribe the first N videos of ":

1. Slugify the channel name → kebab-case (e.g. Two Minute Papers → two-minute-papers).
2. Create the folder channels/<slug>/transcripts/.
3. List the channel's videos (newest first by default unless the user specifies otherwise) and pick the first N (default 20).
4. For each video, fetch a timestamped transcript using:

   python3 ~/.hermes/skills/media/youtube-content/scripts/fetch_transcript.py \
     "<VIDEO_URL>" --text-only --timestamps
   

   Save it to channels/<slug>/transcripts/<NN>-<video-title-slug>.md where NN is a zero-padded index (01..20).
5. Each transcript file should start with a small frontmatter block:

   # <Video Title>
   
   - **Channel:** <Channel Name>
   - **URL:** <Video URL>
   - **Published:** <YYYY-MM-DD>
   - **Duration:** <HH:MM:SS>
   
   ---
   
   <timestamped transcript here>
   

6. Write a channels/<slug>/README.md with the channel name, URL, and a numbered index of the videos (title + link).
7. Update the "Channels" table in this root README.md.
8. Commit with a message like Add <channel-name>: <N> video transcripts and push.

Conventions

• One commit per channel batch (not per video).
• Filenames: lowercase, kebab-case, ASCII only, max ~80 chars before the .md.
• If a video has no available transcript, write a stub file noting the reason and continue — don't fail the batch.
• Skip private/members-only/age-restricted videos and note them in the channel README.

Relevant Skills

• media/youtube-content — transcript fetcher + channel listing helpers
• devops/gitea-private-server — pushing to this Forgejo remote