---

name: semantic-search description: Natural language code search, pattern detection, semantic code analysis version: 1.0.0 tags: [search, semantic, nlp, code-search, pattern-detection]

Semantic Search Skill

Purpose

The Semantic Search Skill enables natural language searching of codebases, detecting patterns, finding similar code, and understanding code semantics. It goes beyond text matching to understand code meaning and relationships.

Key Capabilities:

• Natural language code search ("find database connection functions")
• Semantic similarity detection
• Code pattern recognition
• Cross-reference analysis
• API usage discovery
• Duplicate code detection

Target Token Savings: 75% (from ~2600 tokens to ~650 tokens)

When to Use

• Finding code by description
• Discovering similar implementations
• Locating API usage patterns
• Detecting code duplication
• Understanding code relationships
• Refactoring analysis

Operations

1. search

Natural language code search.

2. find-similar

Finds semantically similar code.

3. detect-patterns

Detects common code patterns.

4. find-duplicates

Identifies duplicate or similar code blocks.

5. analyze-usage

Analyzes API usage patterns.

Scripts

Copy# Natural language search
python ~/.claude/skills/semantic-search/scripts/main.py \
  --operation search \
  --query "database connection functions" \
  --path ./src

# Find similar code
python ~/.claude/skills/semantic-search/scripts/main.py \
  --operation find-similar \
  --file ./src/utils.py \
  --function process_data

# Detect patterns
python ~/.claude/skills/semantic-search/scripts/main.py \
  --operation detect-patterns \
  --path ./src

# Find duplicates
python ~/.claude/skills/semantic-search/scripts/main.py \
  --operation find-duplicates \
  --path ./src \
  --threshold 0.8

Configuration

Copy{
  "semantic-search": {
    "search": {
      "max_results": 10,
      "similarity_threshold": 0.7,
      "include_comments": true
    },
    "patterns": {
      "min_occurrences": 3,
      "min_pattern_lines": 5
    },
    "duplicates": {
      "similarity_threshold": 0.85,
      "min_lines": 10
    }
  }
}

Examples

Example 1: Natural Language Search

Copypython ~/.claude/skills/semantic-search/scripts/main.py \
  --operation search \
  --query "error handling functions" \
  --path ./src

Output:

Copy{
  "success": true,
  "operation": "search",
  "query": "error handling functions",
  "results": [
    {
      "file": "src/utils.py",
      "function": "handle_error",
      "line": 45,
      "relevance": 0.92,
      "snippet": "def handle_error(error, context):"
    },
    {
      "file": "src/api.py",
      "function": "process_exception",
      "line": 123,
      "relevance": 0.87,
      "snippet": "def process_exception(exc):"
    }
  ],
  "execution_time_ms": 234
}

Example 2: Find Similar Code

Copypython ~/.claude/skills/semantic-search/scripts/main.py \
  --operation find-similar \
  --file ./src/utils.py \
  --function process_data

Output:

Copy{
  "success": true,
  "operation": "find-similar",
  "target": "utils.py:process_data",
  "similar_functions": [
    {
      "file": "src/handlers.py",
      "function": "transform_data",
      "similarity": 0.89,
      "reason": "Similar data processing logic"
    },
    {
      "file": "src/processors.py",
      "function": "format_data",
      "similarity": 0.76,
      "reason": "Common transformation patterns"
    }
  ],
  "execution_time_ms": 345
}

Example 3: Detect Code Patterns

Copypython ~/.claude/skills/semantic-search/scripts/main.py \
  --operation detect-patterns \
  --path ./src

Output:

Copy{
  "success": true,
  "operation": "detect-patterns",
  "patterns_found": 5,
  "patterns": [
    {
      "type": "try_except_pattern",
      "occurrences": 12,
      "description": "Try-except with logging",
      "example_location": "src/api.py:45"
    },
    {
      "type": "context_manager",
      "occurrences": 8,
      "description": "Database connection with context manager",
      "example_location": "src/db.py:23"
    }
  ],
  "execution_time_ms": 456
}

Example 4: Find Duplicate Code

Copypython ~/.claude/skills/semantic-search/scripts/main.py \
  --operation find-duplicates \
  --path ./src \
  --threshold 0.85

Output:

Copy{
  "success": true,
  "operation": "find-duplicates",
  "duplicates_found": 3,
  "duplicates": [
    {
      "files": ["src/utils.py:45-60", "src/helpers.py:123-138"],
      "similarity": 0.94,
      "lines": 15,
      "recommendation": "Extract to shared function"
    },
    {
      "files": ["src/api.py:78-92", "src/handlers.py:45-59"],
      "similarity": 0.87,
      "lines": 14,
      "recommendation": "Consider refactoring to common utility"
    }
  ],
  "execution_time_ms": 567
}

Token Economics

Without Skill: ~2600 tokens With Skill: ~650 tokens (75% savings)

Success Metrics

• Search accuracy: >85% relevance
• Pattern detection: >90% accuracy
• Duplicate detection: >95% precision
• Execution time: <500ms for search

---

Semantic Search Skill v1.0.0 - Understanding code through semantics