Extract Videos from YouTube Channel

Why? Manually tracking YouTube videos is tedious and error-prone. This skill automates extracting video metadata (titles, durations, URLs) into a CSV for systematic transcript downloading and analysis.

Quick Start

# Recurring channel (adds to channels.yaml for future syncs)
mkdir -p data/<channel-name>
transcript-extract https://www.youtube.com/@ChannelName/videos \
  --count 50 \
  --append-csv data/<channel-name>/videos.csv \
  --register-channel

# One-time extraction (conferences, playlists)
mkdir -p data/<name>
transcript-extract <youtube_url> --count 100 --append-csv data/<name>/videos.csv

Workflow

1. Determine Extraction Type

--register-channel

[!TIP] Use

--register-channel

only for channels you want to sync regularly. It adds the channel to

data/channels.yaml

for the

sync-all-channels

skill.

2. Create Directory Structure

mkdir -p data/<channel-name>

Naming conventions:

• Use lowercase with hyphens:

  aws-reinvent-2025

  ,

  veritasium

  ,

  lexfridman

• Match the YouTube handle when possible:

  @veritasium

  becomes

  veritasium

• For conferences, include year:

  pycon-2025

  ,

  kubecon-eu-2024

3. Run Extraction Command

transcript-extract <channel_url> \
  --count <N> \
  --append-csv data/<channel-name>/videos.csv \
  [--register-channel]

Parameters:

--count, -n

--append-csv

--register-channel

channels.yaml

--output, -o

--verbose, -v

[!CAUTION] The

--count

in the command sets the INITIAL extraction count. The

count

in

channels.yaml

(set by

--register-channel

) controls FUTURE sync counts. These are independent values.

4. Verify Extraction

# Check CSV was created with expected columns
head -3 data/<channel-name>/videos.csv

# Count extracted videos
wc -l data/<channel-name>/videos.csv

Expected CSV columns:

• url

  - Full YouTube video URL

• title

  - Video title

• duration_minutes

  - Video length

• view_count

  - Number of views

• description

  - Video description

• transcript_downloaded

  - Tracking field (initially empty)

• summary_done

  - Tracking field (initially empty)

Examples

Adding a New Channel for Regular Syncing

# 1. Create directory
mkdir -p data/veritasium

# 2. Extract initial batch with registration
transcript-extract https://www.youtube.com/@veritasium/videos \
  --count 100 \
  --append-csv data/veritasium/videos.csv \
  --register-channel

# Result: 100 videos extracted, channel added to channels.yaml with count: 100

One-Time Conference Extraction

# Extract AWS re:Invent talks (won't be synced later)
mkdir -p data/aws-reinvent-2025
transcript-extract https://www.youtube.com/@AWSEventsChannel/videos \
  --count 200 \
  --append-csv data/aws-reinvent-2025/videos.csv

# No --register-channel = not added to channels.yaml

Re-running on Existing CSV (Incremental Update)

# Running again only adds NEW videos (duplicates auto-skipped)
transcript-extract https://www.youtube.com/@veritasium/videos \
  --count 20 \
  --append-csv data/veritasium/videos.csv

# Safe to run multiple times - existing videos preserved

Troubleshooting

https://www.youtube.com/@ChannelName/videos

/videos

mkdir -p data/<name>

@handle/videos

channels.yaml

--register-channel

channels.yaml

--count

channels.yaml

Common Mistakes

1. Forgetting

   /videos

   in URL
   • Wrong:

     https://www.youtube.com/@veritasium

   • Right:

     https://www.youtube.com/@veritasium/videos

2. Using

   --register-channel

   for one-time extractions
   • This pollutes

     channels.yaml

     with channels you won't sync
   • Only use for channels you want in the regular sync rotation

3. Expecting transcripts to download

   • This skill ONLY extracts video metadata to CSV
   • Use

     download-transcripts

     skill to actually fetch transcripts

4. Not creating directory first

   • The

     --append-csv

     path requires parent directory to exist
   • Always run

     mkdir -p data/<name>

     before extraction

5. Confusing command

   --count

   with

   channels.yaml

   count
   • Command

     --count

     : How many videos to extract NOW

   • channels.yaml

     count

     : How many videos for FUTURE syncs
   • Initial extraction might be 200, but sync count might be 30

Quality Checklist

Before considering extraction complete:

• Directory created in

  data/

  with proper naming
• CSV exists with expected video count
• CSV has all required columns (url, title, duration_minutes, etc.)
• If recurring channel: entry exists in

  channels.yaml

• If recurring channel:

  channels.yaml

  count is set appropriately for future syncs

Next Steps

After extracting videos:

1. Download transcripts: Use the

   download-transcripts

   skill
2. Sync channels later: Use

   sync-all-channels

   for registered channels
3. Add summaries: Use

   summarize-transcripts

   after downloading