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:

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

Top 3-5 Quick Wins: Highest ROI (frequency × impact)
Strategic Skills: High impact even if lower frequency
Quick Automations: High frequency, simpler to implement
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:

[Date]: "[Brief excerpt]"
[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