transcript-learnlisted

# /transcript-learn — Video Knowledge Ingestion Extract structured knowledge from video transcripts and integrate into DQIII8. ## Usage ``` /transcript-learn <youtube-url> /transcript-learn <path-to-transcript.txt> /transcript-learn --batch <file-with-urls.txt> ``` ## Pipeline ### Step 1: Get transcript - If YouTube URL: use youtube-transcript MCP tool `get_transcript` (without timestamps for long videos) - If .txt file: read directly - If transcript > 25000 tokens: process in a forked subagent to save context ### Step 2: Analyze content Extract from the transcript: 1. **Key concepts** — main ideas, definitions, frameworks 2. **Techniques/methods** — actionable processes, step-by-step workflows 3. **Tools/technologies** — software, libraries, APIs mentioned 4. **Data/metrics** — numbers, benchmarks, comparisons 5. **Quotes** — notable statements with timestamp if available 6. **Action items** — things we could implement in DQ or projects ### Step 3: Classify domain Map content to DQ domain: - `formal_sciences`: math, logic, algorithms, statistics - `natural_sciences`: biology, physics, chemistry, nutrition - `social_sciences`: finance, marketing, law, economics - `humanities_arts`: writing, philosophy, history, design - `applied_sciences`: programming, devops, web dev, AI/ML ### Step 4: Generate output Based on content type, generate ONE OR MORE of: **A) Knowledge chunk** (factual knowledge): Save to: `knowledge/{domain}/{topic_slug}.md` Format: use template `t