Intent Classification Skill

Classify intent for: $ARGUMENTS

System Prompt

⚠️ EXECUTION DIRECTIVE: When the user invokes this command, you MUST:

IMMEDIATELY execute - no questions, no explanations first
ALWAYS show full output from script/tool execution
ALWAYS provide summary after execution completes

DO NOT:

Say "I don't need to take action" - you ALWAYS execute when invoked
Ask for confirmation unless requires_confirmation: true in frontmatter
Skip execution even if it seems redundant - run it anyway

The user invoking the command IS the confirmation.

Arguments

$ARGUMENTS - User Request (required)

Specify user prompt to classify:

Market research: "Research Cursor IDE pricing"
Competitive analysis: "Analyze Cursor vs GitHub Copilot"
General research: "Help with competitive analysis"
Vague request: "Need market research"

Default Behavior

Requires user prompt as argument. Returns classification result with:

Detected intent type
Confidence score (0-1)
Recommended agent
Suggested scope
Execution strategy

Description

Advanced natural language intent classification for market research automation. Analyzes user prompts to automatically detect research intent, confidence levels, entities, and optimal execution strategies.

Core Intent Classification Patterns

Market Research Intents (High Confidence ≥85%)

competitive_analysis:
  patterns: ["competitive analysis", "analyze competitors", "competitive landscape", "competitor research"]
  entities: ["company names", "product names"]
  auto_agent: "competitive-market-analyst"
  confidence: 0.90

pricing_research:
  patterns: ["pricing", "pricing strategy", "cost", "subscription", "plans", "price comparison"]
  entities: ["company names", "product tiers"]
  auto_agent: "web-search-researcher"
  confidence: 0.85

market_landscape:
  patterns: ["market landscape", "market analysis", "market overview", "industry analysis"]
  entities: ["market segments", "industry terms"]
  auto_agent: "competitive-market-analyst"
  confidence: 0.88

feature_comparison:
  patterns: ["feature comparison", "features", "capabilities", "functionality comparison"]
  entities: ["product names", "feature categories"]
  auto_agent: "competitive-market-analyst"
  confidence: 0.87

Comparative Analysis Intents (High Confidence ≥80%)

direct_comparison:
  patterns: ["vs", "versus", "compared to", "compare", "A vs B"]
  entities: ["company A", "company B"]
  auto_agent: "competitive-market-analyst"
  confidence: 0.95
  scope: "comparative_analysis"

positioning_analysis:
  patterns: ["positioning", "market position", "how to position", "competitive positioning"]
  entities: ["company names", "target markets"]
  auto_agent: "competitive-market-analyst" 
  confidence: 0.82
  scope: "strategic_positioning"

Workflow Complexity Intents (Medium Confidence 60-79%)

comprehensive_research:
  patterns: ["comprehensive", "complete analysis", "full research", "in-depth", "thorough"]
  auto_agent: "orchestrator"
  confidence: 0.75
  workflow: "multi_agent_research"

quick_overview:
  patterns: ["quick", "brief", "overview", "summary", "fast research"]
  auto_agent: "web-search-researcher"
  confidence: 0.70
  scope: "summary_research"

Entity Detection Algorithms

Company/Product Detection

def detect_entities(text):
    companies = [
        "Cursor", "GitHub Copilot", "Tabnine", "Codeium", "Replit", 
        "Amazon CodeWhisperer", "JetBrains AI", "Sourcegraph Cody",
        "OpenAI Codex", "Anthropic Claude", "Google Bard"
    ]
    
    products = [
        "IDE", "code assistant", "development tool", "AI IDE", 
        "coding assistant", "code completion", "pair programming"
    ]
    
    detected_companies = [c for c in companies if c.lower() in text.lower()]
    detected_products = [p for p in products if p.lower() in text.lower()]
    
    return {
        "companies": detected_companies,
        "products": detected_products,
        "has_comparison": any(word in text.lower() for word in ["vs", "versus", "compared to", "compare"])
    }

Scope Keywords Detection

def detect_scope_keywords(text):
    scope_categories = {
        "pricing": ["pricing", "cost", "price", "subscription", "plans", "tiers", "pricing strategy"],
        "features": ["features", "capabilities", "functionality", "comparison", "technical"],
        "positioning": ["positioning", "market position", "strategy", "differentiation"],
        "market": ["market", "landscape", "overview", "analysis", "size", "trends"],
        "technical": ["technical", "architecture", "implementation", "API", "integration"]
    }
    
    detected_scopes = []
    for category, keywords in scope_categories.items():
        if any(keyword in text.lower() for keyword in keywords):
            detected_scopes.append(category)
    
    return detected_scopes

Confidence Scoring Algorithm

Multi-Factor Confidence Calculation

def calculate_confidence(text, intent_patterns):
    confidence_factors = {
        "exact_pattern_match": 0.4,      # Exact phrase found
        "keyword_density": 0.2,          # Relevant keywords per total words
        "entity_specificity": 0.2,       # Specific companies/products mentioned
        "context_clarity": 0.1,          # Clear vs ambiguous phrasing
        "scope_specificity": 0.1         # Specific scope keywords present
    }
    
    base_confidence = calculate_pattern_match(text, intent_patterns)
    
    # Boost for entity specificity
    entities = detect_entities(text)
    if entities["companies"]:
        base_confidence += 0.1
    
    # Boost for clear scope
    scopes = detect_scope_keywords(text)
    if len(scopes) >= 2:
        base_confidence += 0.05
    
    # Boost for comparison indicators
    if entities["has_comparison"]:
        base_confidence += 0.15
    
    return min(base_confidence, 1.0)

Auto-Execution Decision Matrix

Confidence-Based Routing

execution_strategy:
  high_confidence: # ≥80%
    action: "auto_execute"
    message: "🎯 Auto-detected: {intent} - Executing {agent}..."
    
  medium_confidence: # 40-79%
    action: "quick_confirmation" 
    message: "📊 Detected: {intent}. Proceed with {scope}? (y/n/modify)"
    
  low_confidence: # <40%
    action: "clarifying_questions"
    message: "🤔 I can help with: {options}. What's your focus?"

Smart Default Generation

def generate_smart_defaults(intent, entities, scopes):
    defaults = {
        "agent": select_optimal_agent(intent, scopes),
        "scope": prioritize_scope_areas(scopes),
        "entities": focus_entity_list(entities),
        "methodology": select_methodology(intent, entities)
    }
    
    return defaults

Enhanced Prompt Generation

Context-Aware Prompt Building

def build_enhanced_prompt(original_prompt, classification_results):
    enhanced_prompt = f"""
Original Request: {original_prompt}

Auto-Detected Context:
- Intent: {classification_results.intent} (confidence: {classification_results.confidence})
- Focus Areas: {classification_results.scopes}
- Key Entities: {classification_results.entities}
- Recommended Approach: {classification_results.methodology}

Enhanced Instructions:
{generate_contextual_instructions(classification_results)}

Execute this research with your enhanced context awareness, providing progress updates at your defined checkpoints.
"""
    return enhanced_prompt

Usage Integration Examples

Example 1: High Confidence Auto-Execution

User Input: "Research Cursor vs GitHub Copilot pricing for enterprise customers"

Classification Results:
- Intent: pricing_research + direct_comparison
- Confidence: 0.95 (HIGH)
- Entities: ["Cursor", "GitHub Copilot"] 
- Scope: ["pricing", "enterprise"]
- Agent: competitive-market-analyst

Auto-Execution:
🎯 Auto-detected: Enterprise pricing comparison - Launching competitive-market-analyst...

Example 2: Medium Confidence Quick Confirmation

User Input: "I need to understand the competitive landscape"

Classification Results:
- Intent: competitive_analysis
- Confidence: 0.65 (MEDIUM)
- Entities: []
- Scope: ["competitive", "market"]
- Agent: competitive-market-analyst

Quick Confirmation:
📊 Detected: Competitive landscape analysis for AI development tools.
Proceed with comprehensive competitor research? (y/n/modify scope)

Example 3: Low Confidence Clarification

User Input: "Help me with research"

Classification Results:
- Intent: unclear
- Confidence: 0.25 (LOW)
- Entities: []
- Scope: []

Clarifying Questions:
🤔 I can help with several types of research:
[a] Competitive market analysis
[b] Pricing strategy research  
[c] Technical feature comparison
[d] Market landscape overview

What's your primary focus?

Implementation Integration

Command Integration

This skill integrates with /smart-research command:

/smart-research "user prompt here"
# Automatically applies intent classification and executes optimal workflow

Agent Enhancement Integration

Enhanced agents use classification results:

agent_context_enhancement:
  - Receives classification results in enhanced prompt
  - Auto-adapts methodology based on detected intent/scope
  - Provides targeted progress reporting
  - Suggests relevant expansion options

Success Metrics

Classification Accuracy

Target: 85% correct intent classification
Measurement: User confirmation/correction rates
Optimization: Continuous learning from user feedback

Automation Efficiency

Target: 70% reduction in required user prompting
Measurement: Prompt count before/after implementation
Optimization: Confidence threshold tuning

User Satisfaction

Target: 90% user preference for automated vs manual
Measurement: User feedback and usage patterns
Optimization: Balance automation vs control preferences