YouTube Video Review Skill

Purpose

Extract actionable business insights from YouTube videos when users share URLs or transcripts. Focus on implementation steps, not just summaries.

Trigger Conditions

•User shares a YouTube URL (youtube.com/watch, youtu.be)
•User pastes video transcript or summary
•User asks "how do we leverage this" or "what can we learn from this"
•User mentions reviewing a video for business insights

Workflow

1. Content Acquisition

Since YouTube cannot be directly fetched, request one of:

•User-provided transcript (from YouTube's transcript feature)
•User-provided summary or key points
•Screenshot of video content
•Third-party transcript service output

2. Analysis Framework

Extract insights in this structure:

markdown

## Video Summary
[1-2 sentence overview]

## Key Takeaways
1. [Actionable insight #1]
2. [Actionable insight #2]
3. [Actionable insight #3]

## Implementation for [User's Business]
### Quick Wins (This Week)
- [ ] Step 1
- [ ] Step 2

### Medium-Term (This Month)
- [ ] Step 1
- [ ] Step 2

### Strategic (This Quarter)
- [ ] Step 1
- [ ] Step 2

## Tools/Resources Mentioned
- Tool 1: [What it does] - [URL if available]
- Tool 2: [What it does] - [URL if available]

## Warnings/Gotchas
- [Any risks or caveats mentioned]

3. Business Context Mapping

Always map insights to the user's specific context:

•Storage Scout: Storage arbitrage, Flutter app, GitHub automation
•AI Automation: Claude Code workflows, agent patterns
•SaaS/Product: Feature development, user acquisition

Example Analysis

Input: Nate Herk's "Clawdbot vs Claude Code" video

Output:

markdown

## Video Summary
Comparison of Clawdbot (autonomous AI agent) vs Claude Code for developer productivity.
Claude Code wins for coding; Clawdbot wins for ambient assistant tasks.

## Key Takeaways
1. Claude Code provides 10-30% coding speedup with proper prompts
2. Autonomous agents need strict security isolation (separate VPS, limited API keys)
3. Start with read-only access, graduate to write access after trust is established

## Implementation for Storage Scout
### Quick Wins (This Week)
- [ ] Create Claude Code playbook with standard prompts for Flutter/Dart
- [ ] Document project context in CLAUDE.md (already done!)
- [ ] Set up budget alerts on Anthropic API usage

### Medium-Term (This Month)
- [ ] Build CLI workflow for "storage market audit" (fetch prices → analyze → report)
- [ ] Add automated security scan to CI pipeline

### Strategic (This Quarter)
- [ ] Evaluate Clawdbot for ambient tasks (email triage, calendar management)
- [ ] Build "Storage Scout as a Service" for other arbitrageurs

Integration Points

With RLHF System

After analysis, prompt for feedback:

code

Was this analysis helpful for your business?
👍 = actionable insights
👎 = too generic or missed the point

With Session State

Save analysis to .claude/RALPH_STATE.md for continuity:

markdown

## Last Video Analyzed
- URL: [youtube link]
- Date: [timestamp]
- Key Action: [most important next step]

Commands

Trigger Skill

code

/youtube-review [URL or "analyze the video I just shared"]

Export Analysis

Analysis can be exported to:

•GitHub Issue (for tracking implementation)
•Notion page (via API)
•Markdown file in docs/insights/

Limitations

•Cannot directly fetch YouTube content (use transcripts)
•Cannot verify video claims (user should fact-check)
•Analysis quality depends on transcript quality