11 KiB

Raw Blame History

Extract Patterns from Iterations

Analyze generated iterations to extract successful patterns for the pattern library.

Usage

/project:extract-patterns <iterations_dir> <pattern_library_path> [analysis_depth]

Arguments

iterations_dir - Directory containing generated iterations to analyze
pattern_library_path - Path where pattern library JSON will be saved
analysis_depth - Optional: "quick" (top 3 patterns) or "deep" (top 5 patterns, default)

Examples

# Extract patterns from output directory
/project:extract-patterns output pattern_library/patterns.json

# Quick extraction (3 patterns per category)
/project:extract-patterns output pattern_library/patterns.json quick

# Deep analysis (5 patterns per category)
/project:extract-patterns output pattern_library/patterns.json deep

How It Works

This command implements pattern recognition inspired by multi-shot prompting principles:

Example Collection: Gather all iterations as potential examples
Quality Scoring: Evaluate each iteration across multiple dimensions
Pattern Identification: Extract successful approaches and techniques
Example Selection: Choose 3-5 most exemplary and diverse patterns
Library Update: Save patterns in structured format for future use

Implementation Steps

You are the pattern extraction agent. Follow this workflow:

Step 1: Load and Inventory Iterations

# List all files in iterations directory
Bash: find iterations_dir -type f | sort

# Read each iteration file
For each file:
  - Read file
  - Store content
  - Note file path and metadata

Step 2: Analyze Structural Patterns

Extract patterns related to file organization and architecture:

For each iteration:
  Analyze:
  - File structure and organization
  - Naming conventions used
  - Code/content architecture
  - Module organization (if applicable)
  - Separation of concerns

Score based on:
  - Clarity and consistency
  - Scalability of approach
  - Adherence to best practices
  - Innovation in structure

Identify top 3-5 structural patterns:

{
  "name": "Modular Three-Layer Architecture",
  "description": "Separates data, logic, and presentation into distinct sections",
  "example_file": "output/iteration_7.html",
  "key_characteristics": [
    "Clear section boundaries with comments",
    "Data defined separately from rendering logic",
    "Reusable component structure",
    "Self-documenting organization"
  ],
  "success_metrics": "High readability score (95%), easy to extend, follows separation of concerns",
  "code_snippet": "<!-- Example of clear section separation -->\n<!-- DATA LAYER -->\n...\n<!-- LOGIC LAYER -->\n...\n<!-- PRESENTATION LAYER -->\n..."
}

Step 3: Analyze Content Quality Patterns

Extract patterns related to content excellence:

For each iteration:
  Analyze:
  - Documentation quality and completeness
  - Code/content clarity and readability
  - Comment quality and usefulness
  - Error handling approaches
  - User experience considerations

Score based on:
  - Comprehensiveness of documentation
  - Clarity of explanations
  - Thoughtfulness of implementation
  - Attention to edge cases

Identify top 3-5 content quality patterns:

{
  "name": "Progressive Disclosure Documentation",
  "description": "Layers documentation from overview to deep technical details",
  "example_file": "output/iteration_12.html",
  "key_characteristics": [
    "High-level summary at top",
    "Inline comments for complex logic",
    "Detailed API documentation in separate section",
    "Examples embedded with explanations"
  ],
  "success_metrics": "Easy for beginners and experts alike, 100% of functions documented",
  "code_snippet": "/**\n * HIGH-LEVEL: This function renders...\n * \n * TECHNICAL: Uses D3.js force simulation...\n * \n * EXAMPLE: renderGraph(data) -> visual output\n */"
}

Step 4: Analyze Innovation Patterns

Extract creative and novel approaches:

For each iteration:
  Analyze:
  - Unique problem-solving approaches
  - Creative implementations
  - Novel feature combinations
  - Innovative UX/DX decisions
  - Unexpected but effective solutions

Score based on:
  - Originality compared to other iterations
  - Effectiveness of the innovation
  - Replicability in other contexts
  - Impact on quality or functionality

Identify top 3-5 innovation patterns:

{
  "name": "Self-Validating Data Pipeline",
  "description": "Data includes validation logic that runs automatically",
  "example_file": "output/iteration_15.html",
  "key_characteristics": [
    "Data objects include .validate() method",
    "Automatic validation before rendering",
    "Clear error messages for invalid data",
    "Self-documenting data requirements"
  ],
  "success_metrics": "Zero runtime errors due to data issues, excellent developer experience",
  "code_snippet": "const dataPoint = {\n  value: 42,\n  validate() {\n    if (this.value < 0) throw new Error('...');\n    return true;\n  }\n};"
}

Step 5: Analyze Quality & Testing Patterns

Extract patterns for ensuring quality:

For each iteration:
  Analyze:
  - Testing approaches (if present)
  - Validation strategies
  - Error handling patterns
  - Defensive programming techniques
  - Quality assurance methods

Score based on:
  - Robustness of error handling
  - Thoroughness of validation
  - Testability of implementation
  - Resilience to edge cases

Identify top 3-5 quality patterns:

{
  "name": "Guard Clause Pattern with Fallbacks",
  "description": "Early validation with graceful degradation for missing data",
  "example_file": "output/iteration_9.html",
  "key_characteristics": [
    "Input validation at function entry",
    "Specific error messages for each validation",
    "Fallback defaults for optional parameters",
    "Never crashes, always renders something"
  ],
  "success_metrics": "100% uptime even with malformed data, excellent error messages",
  "code_snippet": "function render(data) {\n  if (!data) return renderEmpty();\n  if (!Array.isArray(data)) data = [data];\n  if (data.length === 0) return renderNoData();\n  // ... continue with rendering\n}"
}

Step 6: Build Pattern Library JSON

Construct the complete pattern library:

{
  "version": "1.2",
  "last_updated": "2025-10-10T14:30:00Z",
  "total_iterations_analyzed": 15,
  "analysis_depth": "deep",
  "patterns": {
    "structural": [
      { "name": "...", "description": "...", ... },
      { "name": "...", "description": "...", ... },
      { "name": "...", "description": "...", ... }
    ],
    "content": [
      { "name": "...", "description": "...", ... },
      { "name": "...", "description": "...", ... },
      { "name": "...", "description": "...", ... }
    ],
    "innovation": [
      { "name": "...", "description": "...", ... },
      { "name": "...", "description": "...", ... },
      { "name": "...", "description": "...", ... }
    ],
    "quality": [
      { "name": "...", "description": "...", ... },
      { "name": "...", "description": "...", ... },
      { "name": "...", "description": "...", ... }
    ]
  },
  "metadata": {
    "extraction_date": "2025-10-10T14:30:00Z",
    "source_directory": "output/",
    "iterations_count": 15,
    "patterns_extracted": 12,
    "avg_quality_score": 8.4,
    "most_common_theme": "Modular architecture with clear separation"
  }
}

Step 7: Save and Report

# Write pattern library to JSON file
Write pattern_library_path with JSON content

# Generate extraction report
Create summary showing:
- Patterns extracted per category
- Quality score distribution
- Most innovative iteration
- Most structurally sound iteration
- Recommended patterns for next wave

Pattern Selection Criteria

When choosing which patterns to include (3-5 per category):

Diversity: Select patterns that represent different approaches
Clarity: Choose patterns that are easy to understand and replicate
Effectiveness: Prioritize patterns with demonstrated success
Transferability: Pick patterns applicable to various contexts
Exemplary Quality: Select from top 20% of iterations only

Multi-Shot Prompting Principles Applied

This extraction process implements key multi-shot prompting concepts:

Example Quality: Only top 20% iterations become examples (high bar)
Diversity: 3-5 patterns prevent overfitting to single approach
Relevance: Patterns are categorized for targeted application
Edge Cases: Innovation category captures unusual but effective approaches
Uniform Structure: All patterns follow consistent JSON schema

Update Strategy

If pattern library already exists:

1. Load existing library
2. Extract patterns from NEW iterations only
3. Merge with existing patterns:
   - Keep patterns with highest success metrics
   - Remove duplicates (similar patterns)
   - Maintain 3-5 patterns per category limit
   - Increment version number
   - Update metadata

Validation

Before saving pattern library:

Validate that:
- JSON is well-formed
- Each pattern has all required fields
- Code snippets are valid (if applicable)
- Success metrics are specific and measurable
- Examples are diverse within each category
- Version number is incremented correctly

Output Report

Generate a summary report:

# Pattern Extraction Report

## Analysis Summary
- Iterations analyzed: {count}
- Analysis depth: {quick|deep}
- Patterns extracted: {total}

## Patterns by Category

### Structural Patterns ({count})
1. {pattern_name}: {brief_description}
2. {pattern_name}: {brief_description}
...

### Content Quality Patterns ({count})
1. {pattern_name}: {brief_description}
2. {pattern_name}: {brief_description}
...

### Innovation Patterns ({count})
1. {pattern_name}: {brief_description}
2. {pattern_name}: {brief_description}
...

### Quality & Testing Patterns ({count})
1. {pattern_name}: {brief_description}
2. {pattern_name}: {brief_description}
...

## Exemplary Iterations
- Best structural: {file_path}
- Best content: {file_path}
- Most innovative: {file_path}
- Highest quality: {file_path}

## Pattern Library Saved
Location: {pattern_library_path}
Version: {version}

## Recommendations
- Use {pattern_name} for structural consistency
- Apply {pattern_name} for content quality
- Consider {pattern_name} for innovation
- Implement {pattern_name} for robustness

Notes

Pattern extraction is automatic but can be manually refined
Library grows with each wave but maintains size limit (3-5 per category)
Patterns serve as multi-shot examples for future iterations
Quality bar rises naturally as better patterns are discovered
Pattern library is spec-agnostic and can be reused across projects

/project:infinite-synthesis - Main loop using pattern library
/project:analyze-patterns - Analyze pattern library effectiveness

11 KiB Raw Blame History