Learning Sessions Skill

Overview

System for extracting, analyzing, and storing insights from YouTube videos and other learning content.

Components

1. YouTube Transcript Extraction

Workflow: .github/workflows/youtube_transcript.yml
Trigger: Manual dispatch or Claude-triggered
Output: Transcript JSON + Supabase logging

2. Fallback Strategy

When transcripts are unavailable (no captions, Hebrew videos, etc.):

1. Claude uses noembed.com API to get video title/channel
2. Claude uses web search to find content about the topic
3. Claude analyzes based on available information
4. User can provide manual context if needed

3. Storage Locations

Location	Purpose
skills/learning-sessions/summaries/	Full analysis files
skills/learning-sessions/LEARNINGS.md	Quick reference consolidated
Supabase insights table	Searchable metadata

4. Analysis Template

## [Video Title]
**Source:** [URL]
**Channel:** [Channel Name]
**Duration:** [X minutes]
**Date Watched:** [Date]

### Key Takeaways
1. [Point 1]
2. [Point 2]

### Application to My Business
- **BidDeed.AI:** [How this applies]
- **Life OS:** [How this applies]

### Actionable Items
- [ ] [Action 1]

### Tags
#tag1 #tag2

Usage

From Claude.ai:

1. Share YouTube URL
2. Claude extracts metadata + searches for context
3. Claude provides analysis
4. Say "Log this" to save important sessions

Manual Trigger:

1. Go to: https://github.com/breverdbidder/life-os/actions/workflows/youtube_transcript.yml
2. Click "Run workflow"
3. Enter video URL
4. Download transcript artifact (if available)

Known Limitations

• Videos without captions return "Transcript unavailable"
• Hebrew videos may not have auto-generated captions
• Private/unlisted videos cannot be accessed
• Claude.ai environment has network restrictions; GitHub Actions can access YouTube

Workarounds

For videos without transcripts:

1. Watch the video manually
2. Share key points with Claude
3. Claude will structure and store the learning session