Workflow Pattern Analyzer Web

Workflow Pattern Analyzer (Web Compatible)

Instructions

This skill provides comprehensive conversation pattern analysis using Claude's native chat history tools (recent_chats and conversation_search) to identify skill-worthy automation opportunities with statistical rigor.

Core Capabilities:

• Web interface compatible (no exports required)
• Statistical pattern validation and scoring
• Frequency analysis and temporal tracking
• Evidence-based skill recommendations
• Complete skill package generation

Compatible with: Claude.ai web interface, Claude Code, API

How Analysis Works:

• No scripts or Python files: This is a pure prompt-based analysis using Claude's native capabilities
• Full content analysis: Examines complete conversation content, messages, and patterns (not just titles or names)
• Thread names: Renaming conversations has minimal impact - analysis focuses on actual message content and patterns
• Domain discovery: Categories emerge from your actual usage data, not forced into predefined buckets
• Data-driven approach: Identifies YOUR specific patterns (recipes, image prompting, game design, etc.) rather than assuming business/coding focus

Analysis Framework

Phase 1: Data Collection Strategy

Determine Analysis Scope:

Ask user: "How deep should I analyze your conversation history?"

Options:

• Quick Scan (20-30 conversations, ~2-3 min): Recent patterns and immediate opportunities
• Standard Analysis (50-75 conversations, ~5-7 min): Comprehensive pattern detection
• Deep Dive (100+ conversations, ~10-15 min): Full workflow mapping with temporal trends
• Targeted Search (variable): Focus on specific topics or time periods

Data Collection Process:

1. Broad Sampling: Use recent_chats(n=30) multiple times with varied parameters to get diverse coverage
2. Temporal Distribution: Sample conversations across different time periods (recent, 1 week ago, 1 month ago)
3. Topic Exploration: Use conversation_search for domains mentioned by user or detected in initial sampling
4. Depth vs Breadth: Balance comprehensive coverage with processing efficiency

Phase 2: Pattern Discovery & Classification

Extract patterns using these detection methods:

A. Explicit Pattern Markers

• Repeated phrasing: "format this as...", "make it more...", "apply X style"
• Consistent request structures: "create a [X] that does [Y]"
• Recurring formatting instructions: tables, bullet lists, specific structures
• Tone/voice adjustments: "more casual", "add enthusiasm", "formal version"

B. Implicit Workflow Patterns

• Multi-turn conversation structures: Same workflow across different topics
• Iterative refinement sequences: Request → feedback → revision cycles
• Context re-explanation: Same background info provided repeatedly
• Problem-solving approaches: Consistent debugging/analysis methodologies

C. Domain Discovery (Data-Driven)

• Let domains emerge from the data - Do NOT pre-categorize into standard domains
• Topic frequency analysis: Extract actual subject matter from conversations
  • Examples of specialized domains: recipe transcription, cannabis strains, image prompting, game design, book summaries
  • Examples of traditional domains: coding, business strategy, creative writing, data analysis, technical writing
• Task type patterns: Identify the action types that appear (creation, transformation, analysis, troubleshooting, curation, etc.)
• Niche specialization detection: Look specifically for narrow, specialized topics with high engagement
• Cross-domain workflows: Patterns that span multiple topics
• Domain diversity scoring: Reward finding 8-15 distinct domains vs. forcing into 3-4 buckets

CRITICAL: Avoid fitting patterns into predefined categories. Each user's conversation history will have unique domains based on their actual usage.

Terminal Output - Domain Diversity Visualization:

After completing pattern discovery, display an ASCII chart showing domain distribution:

📊 Domain Distribution Analysis

Business & Strategy    ████████████░░░░░░░░ 12 patterns (32%)
Creative & Writing     ██████████░░░░░░░░░░ 10 patterns (27%)
Image Prompting        ████████░░░░░░░░░░░░  8 patterns (22%)
Learning & Education   ████░░░░░░░░░░░░░░░░  4 patterns (11%)
Recipe & Cooking       ██░░░░░░░░░░░░░░░░░░  2 patterns  (5%)
Gaming & Design        █░░░░░░░░░░░░░░░░░░░  1 pattern   (3%)

✅ Domain Diversity: 6 distinct topic areas detected
✅ No predefined categorization - domains emerged from your data

This validates data-driven discovery of diverse patterns.

D. Niche & Specialized Pattern Detection

Explicitly search for underrepresented domains:

• Hobbyist domains: Recipes, cocktails, cannabis, gardening, gaming, fitness, travel planning
• Creative domains: Story writing, worldbuilding, character development, art direction, music composition
• Prompt engineering: Image generation (Midjourney, Stable Diffusion, DALL-E), video generation, AI art workflows
• Learning & education: Book summaries, concept explanations, study guides, teaching materials
• Personal organization: Resume writing, cover letters, personal branding, goal setting
• Entertainment & media: Game design, narrative design, content creation, video scripts
• Wellness & lifestyle: Meal planning, workout routines, meditation guides, habit tracking

Detection strategy:

• Look for concentrated clusters of 5+ conversations on the same narrow topic
• Identify specialized vocabulary/jargon (strain names, recipe terms, art styles, game mechanics)
• Find recurring templates/formats specific to that domain
• Don't dismiss low-frequency patterns if they show high consistency and complexity
• Pay special attention to patterns that appear in conversation titles or search results
• Consider that niche patterns may have lower frequency but higher value due to specialization

Quality indicators for niche patterns:

• Consistent terminology and domain-specific language
• Recurring output formats or structures
• User demonstrates growing expertise over time
• High engagement (longer conversations, multiple refinements)
• Clear workflow or methodology emerging

E. Temporal Patterns

• Weekly/monthly recurring tasks: Reports, summaries, check-ins
• Event-driven patterns: Meeting prep, post-mortems, launches
• Seasonal trends: Quarterly reviews, annual planning
• Frequency trends: Increasing/stable/decreasing over time

Phase 3: Frequency Analysis & Validation

For each identified pattern, calculate:

Occurrence Metrics

• Absolute frequency: Total instances found in analyzed conversations
• Relative frequency: Percentage of conversations containing pattern
• Temporal distribution: First occurrence, most recent, clustering
• Consistency score: Similarity across pattern instances (0-100%)

Statistical Validation

• Significance threshold: Pattern must appear in >5% of conversations OR >3 absolute instances
• Consistency requirement: 70%+ similarity in requirements/structure across instances
• Sample size consideration: Adjust thresholds based on total conversations analyzed

Evidence Collection

• Extract 2-4 representative conversation excerpts per pattern
• Note variation types (what changes vs what stays constant)
• Document user refinement patterns (common adjustments made)

Phase 4: Skill-Worthiness Scoring (0-10 Scale)

Use extended reasoning to evaluate each pattern across 5 dimensions:

1. Frequency Score (0-10)

• 10: Daily usage (20+ instances or >25% of conversations)
• 8-9: Multiple times per week (10-20 instances or 15-25%)
• 6-7: Weekly usage (5-9 instances or 8-15%)
• 4-5: Bi-weekly to monthly (3-4 instances or 5-8%)
• 2-3: Monthly or less (2 instances or 3-5%)
• 0-1: One-off or <3% of conversations

2. Consistency Score (0-10)

• 10: Identical requirements every time (90-100% similarity)
• 8-9: Highly consistent with minor variations (75-90%)
• 6-7: Core structure consistent, details vary (60-75%)
• 4-5: Recognizable pattern, significant variation (45-60%)
• 2-3: Loosely related, different each time (30-45%)
• 0-1: No discernible consistency (<30%)

3. Complexity Score (0-10)

• 10: Multi-step workflow with decision points, high cognitive load
• 8-9: Complex methodology requiring expertise/frameworks
• 6-7: Moderate complexity with structured approach
• 4-5: Straightforward process with some nuance
• 2-3: Simple task with minimal steps
• 0-1: Trivial one-step operation

4. Time Savings Score (0-10)

• 10: >60 min saved per use (or >10 hours/month total)
• 8-9: 30-60 min per use (or 5-10 hours/month)
• 6-7: 15-30 min per use (or 2-5 hours/month)
• 4-5: 5-15 min per use (or 1-2 hours/month)
• 2-3: 2-5 min per use (or 30-60 min/month)
• 0-1: <2 min per use (<30 min/month)

5. Error Reduction Score (0-10)

• 10: Critical tasks with major error consequences
• 8-9: Common mistakes significantly impact quality
• 6-7: Regular pitfalls that skill could prevent
• 4-5: Occasional errors, modest quality improvement
• 2-3: Minor inconsistencies, small quality gains
• 0-1: No error patterns, quality already consistent

Composite Scoring

• Total Score: Sum of 5 dimensions (0-50 scale)
• Priority Classification:
  • Critical (40-50): Implement immediately
  • High (30-39): Strong candidates for skill creation
  • Medium (20-29): Consider for skill creation
  • Low (10-19): Defer or handle with simple prompts
  • Not Viable (0-9): Not worth skill automation

Phase 5: Relationship Mapping & Consolidation

A. Overlap Detection

• Identify shared components across patterns
• Map overlapping functionality (>40% shared steps)
• Find hierarchical relationships (high-level task composed of sub-tasks)
• Detect sequential workflows (tasks that occur in sequence)

B. Consolidation Strategies

Use extended reasoning to determine:

• Merge (>60% overlap): Combine into single comprehensive skill
• Separate with cross-reference (30-60% overlap): Distinct skills with links
• Hierarchical: Main skill + specialized variants → parent/child structure
• Modular: Extract common elements → shared templates/references

C. Boundary Optimization

Each skill should have:

• Clear purpose: Single, well-defined use case
• Distinct triggers: Easy to know when to use vs other skills
• Minimal overlap: <30% shared functionality with other skills
• Appropriate scope: Not too broad (generic) or narrow (over-specialized)

Phase 6: Prioritization Matrix

Generate 2D matrix visualization:

VALUE/IMPACT (High to Low)
     │
HIGH │  🔥 Quick Wins        ⭐ Strategic
     │  [High-priority         [Complex but
     │   automation]           critical]
     │
     │  ──────────────────────────────
     │
LOW  │  🔧 Automate          ⏸️  Defer
     │  [Nice-to-have         [Not worth
     │   efficiency]           automating]
     │
     └────────────────────────────────
          LOW    FREQUENCY    HIGH

Classify each pattern:

• X-axis: Frequency score (0-10)
• Y-axis: Average of Complexity, Time Savings, Error Reduction (0-10)
• Size indicator: Total composite score
• Color coding: Implementation difficulty

Strategic Recommendations:

1. Top 3-5 Quick Wins: Highest ROI (frequency × impact)
2. Strategic Skills: High impact even if lower frequency
3. Quick Automations: High frequency, simpler to implement
4. Defer List: Patterns not meeting skill-worthiness thresholds

Phase 7: Skill Package Generation

For each approved skill, create:

A. Skill Specification Document

## [Skill Name]

**Pattern Evidence:**
- Frequency: [X instances in Y conversations (Z%)]
- Consistency: [X/10 score]
- Time savings: [X hours/month]

**Composite Score: [X/50]**
- Frequency: [X/10]
- Consistency: [X/10]
- Complexity: [X/10]
- Time Savings: [X/10]
- Error Reduction: [X/10]

**Example Conversations:**
1. [Date]: [Brief excerpt showing pattern]
2. [Date]: [Brief excerpt showing pattern]
3. [Date]: [Brief excerpt showing pattern]

**Pattern Components:**
- **Consistent elements**: [What stays the same]
- **Variable elements**: [What changes per instance]
- **Common refinements**: [Typical adjustments user makes]

**Proposed Skill Structure:**

SKILL.md sections:
1. Overview & trigger conditions
2. [Main workflow methodology]
3. Quality standards
4. Examples

Supporting files needed:
- reference.md: [Detailed framework/methodology]
- templates/: [Reusable output templates]
- examples.md: [Additional use cases]

B. Complete SKILL.md File

Generate production-ready skill with:

• Proper YAML frontmatter (name, description with triggers)
• Clear instructions based on pattern analysis
• Evidence-based examples from actual conversations
• Quality standards derived from user refinement patterns
• Progressive disclosure (link to references for detail)

Output Formats

After analysis completion, present:

Summary Report

# Workflow Pattern Analysis Report
**Analysis Date**: [Timestamp]
**Conversations Analyzed**: [X conversations across Y time period]
**Patterns Identified**: [X patterns]
**Skills Recommended**: [Y skills]

## 📊 Skill Prioritization Matrix

```mermaid
%%{init: {'theme':'base'}}%%
quadrantChart
    title Skill Prioritization: Frequency vs Impact
    x-axis Low Frequency --> High Frequency
    y-axis Low Impact --> High Impact
    quadrant-1 Strategic
    quadrant-2 Quick Wins
    quadrant-3 Defer
    quadrant-4 Automate
    [Skill Name 1]: [freq_score/10, impact_score/10]
    [Skill Name 2]: [freq_score/10, impact_score/10]
    [Skill Name 3]: [freq_score/10, impact_score/10]

```

Legend:

• Quick Wins (top-right): High frequency, high impact - implement first
• Strategic (top-left): Lower frequency but high value - critical capabilities
• Automate (bottom-right): High frequency, simpler - nice efficiency gains
• Defer (bottom-left): Low priority - consider simple prompts instead

🔥 HIGH-PRIORITY OPPORTUNITIES

1. [Skill Name]

Score: [X/50] (Frequency: X/10, Consistency: X/10, Complexity: X/10, Time: X/10, Error: X/10)

Pattern Description: [What you do repeatedly]

Evidence:

• Found in [X] conversations ([Y%] of analyzed sample)
• First seen: [Date], Most recent: [Date]
• Average time per instance: [X minutes]

Example Occurrences:

1. [Date]: "[Brief excerpt]"
2. [Date]: "[Brief excerpt]"

Proposed Skill: "[One-line skill description]"

Time Savings: [X hours/month]

[Repeat for top 5-8 patterns]

💡 MODERATE OPPORTUNITIES

[Briefer summaries of medium-priority patterns]

🎯 QUICK AUTOMATION CANDIDATES

[Simple, high-frequency patterns]

⏸️ DEFERRED PATTERNS

[Patterns that didn't meet skill-worthiness thresholds]

📊 ANALYSIS METADATA

• Total conversations: [X]
• Date range: [earliest] to [latest]
• Unique patterns identified: [X]
• Patterns validated: [Y]
• Cross-pattern overlaps: [Z]
• Recommended consolidations: [N]

### Interactive Follow-Up Options

What would you like to do next?

A. Generate complete SKILL.md files for [top 3-5 skills] B. Deep dive into specific pattern: [skill name] C. Expand analysis with more conversations D. Focus on specific domain/topic area E. Adjust scoring weights and recalculate priorities

## Quality Standards

### Pattern Validation Requirements
- **Minimum frequency**: 3 instances OR >5% of conversations
- **Consistency threshold**: 70% similarity across instances
- **Evidence requirement**: 2-3 conversation excerpts minimum
- **Time savings minimum**: >30 min/month cumulative
- **Statistical significance**: Validated against sample size

### Skill Recommendations
- **Maximum 8-10 skills**: Focus on highest ROI patterns
- **Clear differentiation**: Each skill has distinct purpose
- **Evidence-based design**: Skill structure reflects actual usage patterns
- **Practical focus**: Prioritize patterns with real time/quality impact

### Analysis Rigor
- **No generic patterns**: Avoid broad categories like "writing" or "analysis"
- **Domain diversity validated**: Patterns should span 6+ distinct topic areas (not just business/coding)
- **Niche specializations identified**: Don't dismiss specialized patterns as outliers
- **No artificial categorization**: Avoid forcing patterns into predefined domains (coding, writing, business, analysis)
- **Validated frequencies**: Ensure counts are accurate within sample
- **Temporal awareness**: Note if pattern is emerging, stable, or declining
- **User context**: Consider user's domain expertise and role

## Examples

### Example 1: Quick Scan Request

**User**: "Analyze my recent chats for automation opportunities"

**Claude**:
I'll perform a quick scan of your recent conversation history to identify skill opportunities.

[Uses recent_chats(n=30) multiple times]

📊 **Quick Scan Results** (30 conversations analyzed)

🔥 **TOP OPPORTUNITY**:

**"Email Response Composer"** - Score: 42/50
- Found in 9 conversations (30%)
- Pattern: You regularly ask me to draft professional emails with specific tone and structure requirements
- Time savings: ~90 min/month (10 min per email × 9 emails)
- Evidence: Consistent 3-paragraph structure with greeting, main content, action items

Want me to:
A. Generate complete skill package for this
B. Continue analysis for more patterns (expand to 50-75 conversations)
C. Focus on other domains

### Example 2: Targeted Domain Analysis

**User**: "Find patterns in how I use you for coding tasks"

**Claude**:
I'll analyze your coding-related conversations to identify workflow patterns.

[Uses conversation_search for coding terms + recent_chats filtering]

📊 **Coding Workflow Analysis** (18 relevant conversations)

🔥 **TOP PATTERNS**:

1. **"Code Review Checklist Generator"** - Score: 38/50
   - Found in 7 code review discussions
   - Pattern: You ask for language-specific review checklists before reviewing PRs
   - Consistency: 85% (always includes security, performance, style sections)

2. **"API Documentation Humanizer"** - Score: 35/50
   - Found in 5 documentation tasks
   - Pattern: Converting technical API specs into developer-friendly guides
   - Structure: Always includes quick start, common use cases, troubleshooting

Want me to build skills for these patterns?

### Example 3: Deep Dive Analysis

**User**: "Do a comprehensive analysis of my workflow over the last few months"

**Claude**:
I'll perform a deep analysis of your conversation history. This will take 10-15 minutes to ensure comprehensive coverage.

[Uses recent_chats extensively, conversation_search for discovered topics, temporal sampling]

📊 **Comprehensive Workflow Analysis**
- **Conversations analyzed**: 120
- **Date range**: [3 months]
- **Patterns identified**: 15
- **Skills recommended**: 6

🔥 **CRITICAL PRIORITY** (Score 40-50):

1. **"Weekly Status Report Generator"** - Score: 47/50
   - Frequency: 12 instances (10% of conversations)
   - Consistency: 95% - always same structure
   - Evidence: Every Monday, you format updates in identical 5-section template
   - Time savings: 240 min/month (20 min/week × 4 weeks × 3 months avg)

⭐ **HIGH PRIORITY** (Score 30-39):

2. **"Client Proposal Framework"** - Score: 36/50
3. **"Meeting Notes Structurer"** - Score: 34/50
4. **"Technical Concept Explainer"** - Score: 31/50

[Full analysis report with evidence, prioritization matrix, skill specifications]

**Recommended Implementation Path**:
1. Start with "Weekly Status Report Generator" (highest ROI)
2. Build "Client Proposal Framework" and "Meeting Notes Structurer" next (complementary workflows)
3. Evaluate remaining patterns after 2-4 weeks of usage

Generate complete skill packages now? [Y/N]

## Anti-Patterns to Avoid

**Don't recommend skills for:**
- **One-off variations**: Tasks that seem similar but are fundamentally different each time
- **Over-simplified tasks**: Things easier to just ask directly than invoke a skill
- **Better solved by tools**: When external apps/services do it better
- **Insufficient data**: Patterns with <3 instances or <5% frequency (unless strategic)
- **Generic categories**: Broad skills like "help with writing" or "analyze data"

**Red flags in patterns:**
- High frequency but no consistency (chaotic variation)
- High consistency but very low frequency (use a prompt template instead)
- Pattern is declining over time (user may have found better solution)
- Task requires real-time data or external authentication (needs MCP, not skill)

## When to Use This Skill

**✅ Use this skill when:**
- User requests analysis of their conversation patterns
- User wants to identify automation opportunities
- User asks what skills they should create
- User mentions repetitive tasks or workflows
- User wants evidence-based skill recommendations
- User is in web interface (can't use export-based analysis)

**❌ Don't use this skill when:**
- User has conversation export files available (use export-based plugin instead for more comprehensive analysis)
- User wants cross-platform ChatGPT + Claude analysis (requires exports)
- User has very few conversations (<10) making pattern detection unreliable
- User wants to build specific skill they already have in mind
- User is asking about existing skills or community skills

**⚡ Proactive Use:**
When you detect potential patterns during normal conversation, offer:

💭 Pattern detected: This is the [Xth] time you've asked me to [action].

Would you like me to analyze your conversation history for similar patterns and recommend a Custom Skill? I can identify other automation opportunities you might not have noticed.

[Yes, analyze] [Not now]

## Progressive Disclosure Strategy

**Keep main analysis concise by organizing information hierarchically:**

1. **Quick overview first**: Summary report with top 3-5 opportunities
2. **Details on demand**: Expand specific patterns when user shows interest
3. **Implementation when ready**: Generate complete skill packages only after user approval
4. **Iterative refinement**: Allow user to adjust scoring weights, focus areas, or analysis depth

**Load additional detail only when:**
- User requests deep dive on specific pattern
- Generating complete skill packages (not just analysis)
- User wants to understand scoring methodology in detail
- Building skills for complex domains requiring extensive examples