Hanx Youtube Researcher

Hanx YouTube Researcher Skill

Perform specialized YouTube video analysis using domain-specific prompt templates. This skill extends the base youtube-video-analysis skill with three powerful analysis modes optimized for different video types.

Overview

This skill provides three specialized analysis templates that extract structured, domain-specific insights from YouTube videos:

1. General Analysis - Comprehensive video analysis for any content
2. Trading Strategy Analysis - Structured extraction of trading methodologies
3. Framework/Tool Analysis - Technical documentation extraction from tutorials

Each analysis type uses carefully crafted prompts to extract specific information in JSON format, making the output immediately actionable.

When to Use This Skill

Automatic Triggers

• User provides YouTube URL with "analyze for trading strategy"
• User mentions "extract framework details from video"
• User asks for "structured analysis of video"
• User wants "trading strategy breakdown"

Manual Invocation

# General analysis
python specialized_analyzer.py --url https://youtu.be/example --type general

# Trading strategy analysis
python specialized_analyzer.py --url https://youtu.be/example --type trading

# Framework/tool analysis
python specialized_analyzer.py --url https://youtu.be/example --type framework

Three Analysis Types

1. General Analysis

Best For:

• Educational content
• Conference talks
• Interviews
• Product demos
• General knowledge extraction

Extracts:

• Summary (concise overview)
• Key points (main arguments/ideas)
• Topics covered (list of subjects)
• Insights (notable takeaways)
• Questions answered (what problems this solves)
• References (external sources mentioned)

Output Format: JSON with structured fields

Use Case: "Analyze this Python conference talk and extract the key insights"

2. Trading Strategy Analysis

Best For:

• Trading strategy tutorials
• Market analysis videos
• Technical indicator guides
• Backtesting discussions
• Risk management content

Extracts:

• Strategy name
• Markets (crypto, forex, stocks, etc.)
• Timeframes (1H, 4H, daily, etc.)
• Indicators (RSI, MACD, EMA, etc.)
• Entry conditions (specific rules)
• Exit conditions (specific rules)
• Risk management (position sizing, stops)
• Backtest results (performance data)
• Pros and cons (advantages/disadvantages)
• Key insights (important notes)

Output Format: JSON with trading-specific fields

Use Case: "Extract the trading strategy from this video and give me the entry/exit rules"

3. Framework/Tool Analysis

Best For:

• Framework tutorials
• Tool demonstrations
• Installation guides
• Technical documentation
• Library/package overviews

Extracts:

• Tool/framework name
• Purpose (problem it solves)
• Target users (who should use it)
• Key features (main capabilities)
• Installation process (setup steps)
• Basic usage (getting started)
• Advanced features (power user capabilities)
• Limitations (constraints/drawbacks)
• Alternatives (competing solutions)
• Resources (docs, links, repos)

Output Format: JSON with technical documentation fields

Use Case: "Watch this FastAPI tutorial and create a setup guide with all features listed"

Integration with Existing Skills

Leverages youtube-video-analysis

This skill is a lightweight extension that focuses only on specialized analysis. It delegates to the existing youtube-video-analysis skill for:

• Video downloading (pytubefix)
• Audio extraction (ffmpeg/moviepy)
• Transcription (Whisper AI)

Architecture:

User Request
    ↓
hanx-youtube-researcher (specialized analysis)
    ↓
youtube-video-analysis (transcript extraction)
    ↓
Structured JSON + Markdown Report

Optional RAG Integration

If youtube-rag-storage skill is available, analysis results can be stored in the knowledge base for future retrieval.

Workflow Example

Complete Analysis Flow

# Step 1: User provides video URL
$ python specialized_analyzer.py --url https://youtu.be/dQw4w9WgXcQ --type trading

# Step 2: Tool uses youtube-video-analysis to get transcript
Downloading video...
Extracting audio...
Transcribing with Whisper (base model)...
Transcript ready (5,432 characters)

# Step 3: Tool applies trading strategy prompt template
Analyzing transcript with trading strategy template...
Querying LLM...

# Step 4: Tool outputs structured JSON
{
  "strategy_name": "RSI Divergence Reversal",
  "markets": ["Crypto", "Forex"],
  "timeframes": ["1H", "4H"],
  "indicators": ["RSI (14)", "Volume", "Support/Resistance"],
  "entry_conditions": [
    "RSI shows bullish divergence (higher lows while price makes lower lows)",
    "Price reaches support level",
    "Volume confirmation on reversal candle"
  ],
  "exit_conditions": [
    "RSI reaches overbought (>70)",
    "Price hits resistance level",
    "Stop loss at 2% below entry"
  ],
  "risk_management": [
    "2% maximum risk per trade",
    "1:2 minimum risk/reward ratio"
  ],
  "backtest_results": "67% win rate over 100 trades, 1:2.3 avg R:R",
  "pros": ["Simple to identify", "Works in trending markets"],
  "cons": ["False signals in choppy markets", "Requires discipline"],
  "key_insights": ["Works best on H4 timeframe", "Combine with volume confirmation"]
}

# Step 5: Tool generates markdown report
Saved: output/RSI_Divergence_Reversal_analysis.md
Saved: output/RSI_Divergence_Reversal_analysis.json

Usage Instructions

Basic Usage

from specialized_analyzer import SpecializedAnalyzer

# Initialize analyzer
analyzer = SpecializedAnalyzer()

# Analyze video
result = analyzer.analyze(
    url="https://youtu.be/example",
    analysis_type="general"
)

print(result['summary'])
print(result['key_points'])

Command-Line Usage

# General analysis
python specialized_analyzer.py \
    --url https://youtu.be/example \
    --type general

# Trading analysis with output directory
python specialized_analyzer.py \
    --url https://youtu.be/example \
    --type trading \
    --output ./trading_analysis

# Framework analysis with custom model
python specialized_analyzer.py \
    --url https://youtu.be/example \
    --type framework \
    --model small \
    --output ./framework_docs

Advanced Usage

# Custom LLM provider
result = analyzer.analyze(
    url="https://youtu.be/example",
    analysis_type="trading",
    llm_provider="anthropic",
    llm_model="claude-sonnet-4"
)

# Save to multiple formats
analyzer.save_results(
    result,
    formats=["json", "markdown", "yaml"],
    output_dir="./analysis"
)

# Integrate with RAG
if has_rag_integration:
    analyzer.store_in_rag(result)

Output Formats

JSON Output

{
  "video_id": "dQw4w9WgXcQ",
  "video_url": "https://youtu.be/dQw4w9WgXcQ",
  "analysis_type": "trading",
  "timestamp": "2025-11-01T12:00:00Z",
  "analysis": {
    "strategy_name": "RSI Divergence Reversal",
    "markets": ["Crypto", "Forex"],
    "timeframes": ["1H", "4H"],
    "indicators": ["RSI (14)", "Volume"],
    "entry_conditions": ["..."],
    "exit_conditions": ["..."],
    "risk_management": ["..."],
    "backtest_results": "...",
    "pros": ["..."],
    "cons": ["..."],
    "key_insights": ["..."]
  }
}

Markdown Output

See templates/ folder for formatted markdown templates:

• general_analysis_template.md
• trading_analysis_template.md
• framework_analysis_template.md

Technical Details

Dependencies

Inherits dependencies from youtube-video-analysis:

• pytubefix>=6.0.0 - YouTube downloading
• openai-whisper>=20231117 - Transcription
• moviepy>=1.0.3 - Audio extraction

Additional dependencies:

• requests>=2.31.0 - LLM API calls
• python-dotenv>=1.0.0 - API key management

Architecture

Modular Design:

specialized_analyzer.py (main CLI)
    ↓
analyze_general()      - General analysis
analyze_trading()      - Trading analysis
analyze_framework()    - Framework analysis
    ↓
get_transcript()       - Calls youtube-video-analysis
    ↓
query_llm()           - LLM API integration
    ↓
format_output()       - JSON + Markdown generation

Key Functions:

• analyze_general(url, transcript) - Apply general analysis prompt
• analyze_trading(url, transcript) - Apply trading strategy prompt
• analyze_framework(url, transcript) - Apply framework/tool prompt
• get_transcript(url) - Delegate to youtube-video-analysis
• query_llm(prompt, provider) - Call LLM API with prompt
• format_output(analysis, template) - Generate markdown report

Prompt Templates

The three prompt templates are extracted directly from the hanx YouTube Researcher Agent (lines 109-199 of agent_youtube_researcher.py):

1. GENERAL_PROMPT - Comprehensive video analysis
2. TRADINGSTRATEGYPROMPT - Trading methodology extraction
3. FRAMEWORKTOOLPROMPT - Technical documentation extraction

These prompts are optimized for structured JSON output with specific fields for each analysis type.

Best Practices

Choosing Analysis Type

General Analysis:

• ✅ Use for: Educational content, talks, interviews
• ✅ When: You want broad understanding
• ❌ Avoid: Highly technical content (use framework instead)

Trading Analysis:

• ✅ Use for: Trading strategies, market analysis
• ✅ When: You need specific entry/exit rules
• ❌ Avoid: General financial news (use general instead)

Framework Analysis:

• ✅ Use for: Tool tutorials, installation guides
• ✅ When: You need technical documentation
• ❌ Avoid: Conceptual discussions (use general instead)

Quality Tips

• Clear Audio - Better transcription = better analysis
• Longer Videos - More content = richer analysis
• Focused Content - Single topic videos work best
• Technical Content - Framework/trading types excel here

Performance Optimization

• Model Selection - Use base Whisper model for speed
• Caching - Store transcripts to avoid re-processing
• Batch Processing - Analyze multiple videos in one session
• LLM Provider - Choose provider based on cost/quality needs

Limitations

Current Limitations

• Language - Best results with English content
• Audio Quality - Poor audio affects transcription accuracy
• Video Length - Very long videos (>2 hours) may need chunking
• Specialized Domains - May miss domain-specific jargon

Not Supported

• ❌ Real-time analysis (must download first)
• ❌ Live streams (must be recorded)
• ❌ Age-restricted videos (requires auth)
• ❌ Private videos (requires auth)

Troubleshooting

Common Issues

Issue: "Failed to get transcript"

• Solution: Ensure youtube-video-analysis skill is installed
• Solution: Check video URL is valid and accessible

Issue: "LLM API error"

• Solution: Verify API key is set in environment
• Solution: Check API provider is supported

Issue: "Invalid analysis type"

• Solution: Must be one of: general, trading, framework

Issue: "JSON parsing error"

• Solution: LLM response may not be valid JSON
• Solution: Check raw response in output file

Future Enhancements

Planned Features

• [ ] Support for additional analysis types (security, architecture, etc.)
• [ ] Multi-language support with automatic detection
• [ ] Batch video processing
• [ ] Custom prompt templates via configuration
• [ ] Integration with fstrentspectasks for task generation
• [ ] RAG integration for knowledge persistence

Potential Improvements

• [ ] Automatic analysis type detection based on video content
• [ ] Confidence scores for extracted information
• [ ] Cross-video comparison and aggregation
• [ ] Export to additional formats (YAML, CSV, etc.)

Resources

Documentation

• See reference/prompt_templates.md for full prompt text
• See reference/output_schemas.md for JSON schema definitions
• See examples/ for workflow demonstrations

Example Files

• examples/general_analysis_example.md - General analysis workflow
• examples/trading_analysis_example.md - Trading strategy extraction
• examples/framework_analysis_example.md - Framework documentation

Related Skills

• youtube-video-analysis - Base video processing (required)
• youtube-rag-storage - Knowledge persistence (optional)

---

Version: 1.0.0 Created: 2025-11-01 Source: Hanx YouTube Researcher Agent Maintainer: AI Project Template Team Status: Active Development

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: wrm3
• Source: wrm3/aiproject_template
• License: MIT

Install and usage instructions live in the source repository linked above.