/you2idea-extract

Extract startup ideas from YouTube videos. Two operating modes depending on available tools.

Mode Detection

Check which tools are available:

• With solograph MCP: use source_search, source_list, source_tags, source_related for indexed corpus
• Without MCP (standalone): use yt-dlp + Read for transcript analysis

MCP Tools (if available)

• source_search(query, source="youtube") — semantic search over indexed videos
• source_list() — check indexed video counts
• source_tags() — auto-detected topics with confidence scores
• source_related(video_url) — find related videos by shared tags
• kb_search(query) — cross-reference with knowledge base
• web_search(query) — discover new videos to index

Steps

Mode 1: Index + Analyze (with solograph MCP)

1. Parse input from $ARGUMENTS:

   • URL (https://youtube.com/watch?v=...) → single video index
   • Channel name (GregIsenberg) → channel batch index
   • Query text → search existing corpus (skip to step 4)
   • If empty, ask: “Video URL, channel name, or search query?”

2. Index video(s) via solograph:

   # Install if needed
   pip install solograph  # or: uvx solograph
   
   # Single video
   solograph-cli index-youtube -u "$URL"
   
   # Channel batch (needs web search for discovery)
   solograph-cli index-youtube -c "$CHANNEL" -n 5
   

3. Verify indexing — source_list() to confirm new video count. source_tags() for topic distribution.

4. Search corpus — source_search(query="startup ideas", source="youtube").

5. Cross-reference — kb_search(query) for related existing opportunities (if knowledge base available).

6. Extract insights — for each relevant video chunk:

   • Identify the startup idea mentioned
   • Note timestamp and speaker context
   • Rate idea potential (specificity, market evidence, feasibility)
   • Flag ideas that match trends or validated patterns

7. Write results to docs/youtube-ideas.md or print summary.

Mode 2: Standalone (without MCP)

1. Parse input — same as Mode 1 step 1.

2. Download transcript via yt-dlp:

   # Check yt-dlp is available
   command -v yt-dlp >/dev/null 2>&1 && echo "yt-dlp: ok" || echo "Install: pip install yt-dlp"
   
   # Download subtitles only (no video)
   yt-dlp --write-auto-sub --sub-lang en --skip-download -o "transcript" "$URL"
   
   # Convert VTT to plain text
   sed '/^$/d; /^[0-9]/d; /-->/d; /WEBVTT/d; /Kind:/d; /Language:/d' transcript.en.vtt | sort -u > transcript.txt
   

3. Read transcript — Read the transcript.txt file.

4. Analyze for startup ideas:

   • Scan for business opportunities, pain points, product ideas
   • Note approximate timestamps from VTT cues
   • Rate each idea on specificity and market potential
   • Cross-reference with WebSearch for market validation

5. For channel analysis — download multiple video transcripts:

   # Get video list from channel
   yt-dlp --flat-playlist --print "%(id)s %(title)s" "https://youtube.com/@$CHANNEL" | head -10
   
   # Download transcripts for top videos
   for id in $VIDEO_IDS; do
     yt-dlp --write-auto-sub --sub-lang en --skip-download -o "transcripts/%(id)s" "https://youtube.com/watch?v=$id"
   done
   

6. Write results to docs/youtube-ideas.md with format:

   # YouTube Ideas — [Channel/Video]
   Date: YYYY-MM-DD
   
   ## Idea 1: [Name]
   - **Source:** [Video title] @ [timestamp]
   - **Problem:** [What pain point]
   - **Solution:** [What they propose]
   - **Market signal:** [Evidence of demand]
   - **Potential:** [High/Medium/Low] — [why]
   
   ## Idea 2: ...
   

Common Issues

yt-dlp not found

Fix: pip install yt-dlp or brew install yt-dlp

No subtitles available

Cause: Video has no auto-generated or manual captions. Fix: Try --sub-lang en,ru for multiple languages. Some videos only have auto-generated subs.

solograph MCP not available

Fix: Skills work in standalone mode (yt-dlp + Read). For indexed search across many videos, install solograph: pip install solograph. For enhanced web search, set up SearXNG (private, self-hosted, free).

Too many ideas, hard to prioritize

Fix: Use /validate on the top 3 ideas to score them through STREAM framework.