Symbol Tree Analysis

Use the mcp__symbol-tree__symbol_tree MCP tool to analyze TypeScript codebases.

MANDATORY: Use This Tool FIRST

STOP. Before you use Grep, Glob, or Read to explore code, you MUST call mcp__symbol-tree__symbol_tree first.

This is not optional. The symbol tree tool provides:

• Instant overview of file/module structure
• Function signatures and relationships
• Call graphs showing what calls what
• All without reading hundreds of lines of implementation

This saves tokens and gives better context faster than file searches.

DO NOT

• DO NOT start with Grep to search for patterns - use symbol_tree first
• DO NOT start with Glob to find files - use symbol_tree first
• DO NOT start with Read to understand code - use symbol_tree first
• DO NOT skip this step - it's mandatory for codebase exploration

Correct Workflow

1. FIRST: Call mcp__symbol-tree__symbol_tree to get overview
2. THEN: Use the symbol tree output to identify relevant files/functions
3. ONLY THEN: Use Read/Grep/Glob for specific details if needed

When to Use This Skill

• FIRST when exploring or researching any part of the codebase
• When the user asks "how does X work?" or "where is Y?"
• When understanding code architecture or file structure
• When exploring how functions/classes relate to each other
• When finding where a symbol is defined and what it calls
• When assessing file size/complexity for refactoring

Parameters

• symbol - Function/class name to analyze (e.g., "step", "Runtime")
• file - Specific file to analyze
• depth - Limit call graph depth (default: 3)
• format - "adjacency" (token-efficient) or "tree" (hierarchical view)
• exports_only - Only show exported symbols
• path - Directory to analyze (defaults to cwd)
• text_limit - Maximum output characters (default: 50000)

Examples

• Analyze a specific function: symbol: "step"
• Analyze a file: file: "src/runtime/step.ts"
• Full project overview: path: "src/", exports_only: true
• Disambiguate duplicates: symbol: "step", file: "src/runtime/step.ts"

Instructions

1. Parse the user's request to determine what they want to analyze
2. Call mcp__symbol-tree__symbol_tree with appropriate parameters
3. Present results with a brief explanation:
   • For adjacency format: explain the SYMBOLS and DEPENDENCIES sections
   • For tree format: explain the call hierarchy
   • Highlight interesting patterns (heavily-called functions, external deps, etc.)

Default Parameters

• format: "adjacency" (use this first) - flat list, token-efficient, good for quick overview
• depth: 3 (reasonable default)
• show_files: true
• text_limit: 50000 (allows ~12k tokens of output without truncation)

Only use format: "tree" when you specifically need to see nested call hierarchies. Start with adjacency.

Output Format

Each symbol shows location as: (file:startLine-endLine)

Example: function parse(source: string): Program (parser\parse.ts:10-25)

Using with Read tool:

• offset: Use startLine to begin reading at the symbol
• limit: Use endLine - startLine + 1 to read only the symbol's code

This enables precise code navigation - read exactly the lines you need.