youtube-video-context-analysis

name	YouTube Video Context Analysis
description	Create timestamped YouTube context ledgers that separate transcript evidence, frame evidence, OCR evidence, inference, and uncertainty for extraction, strategy, creative analysis, and claim audits.
version	1.0
format	completion-engine
expert	YouTube Video Context Analysis
domain	video evidence analysis, transcript cleanup, frame sampling, OCR fallback, source-to-skill extraction
workflows	8

YouTube Video Context Analysis

This skill turns a public YouTube URL into a reusable evidence package. It is not a transcript summarizer. It creates a timestamped ledger that keeps observed spoken evidence, observed visual evidence, observed on-screen text, inference, and unavailable evidence in separate lanes.

Core Rule

Never merge inferred visual assumptions into observed evidence. If the system did not capture a frame, OCR text, human visual note, or vision-adapter output, it cannot claim that it saw the visual.

Available Workflows

#	Workflow	Slash Command	Produces
1	Quick Transcript Ledger	/video-transcript-ledger	Metadata, cleaned transcript, spoken-evidence ledger, uncertainty report
2	Full Visual Context Ledger	/video-context-ledger	Transcript plus frame/OCR-aware context package
3	Frame Ledger	/video-frame-ledger	Frame sample plan, extracted frames, frame notes
4	Visual OCR	/video-visual-ocr	OCR notes and on-screen text rows when tooling is available
5	Source-To-Skill Extract	/video-source-extract	Extraction-ready source map for /extract or /extract-forge
6	Context Audit	/video-context-audit	Claim, contradiction, uncertainty, and evidence separation audit
7	Creative Reference Breakdown	/video-creative-reference	Creative reference notes grounded in frames/OCR/transcript
8	Multi-Video Comparison	/video-multi-compare	Cross-video source map and reusable pattern comparison

Primary Local Command

python3 execution/video_context_ledger.py "<youtube-url>" --mode transcript
python3 execution/video_context_ledger.py "<youtube-url>" --mode full
python3 execution/video_context_ledger.py --mode fixture --fixture execution/video_context/fixtures/sample_video_context.json

Output Package

Each run writes to extractions/video-context/<video-id>/:

• metadata.json
• transcript.vtt when subtitles are available
• transcript.txt as the clean continuous transcript
• transcript_segments.json as timestamped spoken-evidence segments
• frames/ when frame extraction is available
• frame-notes.md
• ocr-notes.md
• video-context-ledger.md
• video-context-ledger.json
• analysis.md
• uncertainty-report.md

Source Package Verification

python3 execution/verify_video_context_source_package.py "extractions/video-context/<video-id>"

Use the clean transcript for reading and transcript_segments.json plus video-context-ledger.md for timestamped evidence. Do not treat row-shaped ledger evidence as the human transcript.

Stacking Guide

Stack With	How To Use The Ledger
/extract	Use video-context-ledger.md instead of transcript-only source material
/extract-forge	Use full visual context for rich interviews, lectures, demos, and source-to-skill builds
/research-intelligence-agent	Audit claims with a timestamped evidence map
/creative-design-agent	Convert real visual references into grounded creative direction
/content-media-agent	Build clips, posts, and scripts from timestamped evidence
/adversarial-review	Separate claim, proof, contradiction, and uncertainty

Quick Reference

• Genius context: genius.md
• Source map reference: references/source-map.md
• Quality rubric: references/quality-rubric.md
• Evidence patterns: references/genius-patterns.md
• Hidden knowledge: references/hidden-knowledge.md

Execution Prompts (structure-pure v2)

8 deterministic practitioner prompts — each carries an Output Contract, Output Skeleton, and Quality Gate. When a deliverable matches one, Read it and honor its contract instead of improvising the output shape.

• Context Audit — [video title / id] — skills/youtube-video-context-analysis/references/prompts-v2/context-audit.md
• Creative Reference Breakdown — [video title / id] — skills/youtube-video-context-analysis/references/prompts-v2/creative-reference-breakdown.md
• frame-notes.md — skills/youtube-video-context-analysis/references/prompts-v2/frame-ledger.md
• video-context-ledger.md — skills/youtube-video-context-analysis/references/prompts-v2/full-visual-context-ledger.md
• Multi-Video Comparison — [purpose] — skills/youtube-video-context-analysis/references/prompts-v2/multi-video-comparison.md
• metadata.json — skills/youtube-video-context-analysis/references/prompts-v2/quick-transcript-ledger.md
• Source-to-Skill Extraction Map — [video title / id] — skills/youtube-video-context-analysis/references/prompts-v2/source-to-skill-extraction-map.md
• ocr-notes.md — skills/youtube-video-context-analysis/references/prompts-v2/visual-ocr-notes.md