ast-dependency-analyzer_skill

Overview

This skill analyzes JavaScript and TypeScript codebases to extract exports, imports, and build a complete dependency graph. It uses tree-sitter AST parsing to produce a DependencyContext containing a symbol table, dependency graph, and reverse dependencies. Use it to get a precise, line-numbered map of how modules relate across a project. The output is optimized for programmatic queries and prompt-friendly summaries.

How this skill works

The analyzer parses every .js/.jsx/.mjs/.cjs/.ts/.tsx file with tree-sitter and extracts all export and import nodes including re-exports, type-only imports, dynamic imports, and side-effect imports. It resolves module specifiers where possible to absolute paths and records line numbers, export/import kinds, aliases, and source files. From that data it builds three primary structures: symbol_table (what each file exports), dependency_graph (what each file imports), and reverse_dependencies (who imports each file). Results are cached by content hash and timestamp to speed repeated runs.

When to use it

• Detect barrel files that re-export symbols from other modules
• Find hub files that many modules depend on before refactoring
• Detect circular dependencies and import cycles
• Discover orphaned files that nothing imports
• Analyze import patterns and type vs runtime imports for cleanup

Best practices

• Run analysis from the repository root to get consistent resolved paths
• Use force_refresh when you changed many files to bypass cache
• Combine reverse_dependencies with symbol_table to identify unused exports
• Treat type-only imports separately when making runtime dependency changes
• Exclude generated or build artifacts from the target path to reduce noise

Example use cases

• Barrel detection: find pure and mixed re-export files and rank by consumer count
• Hub detection: identify files with many dependents to prioritize stabilization before changes
• Circular dependency tracing: produce a cycle path for remediation and tests
• Orphan detection: list files with zero consumers for possible deletion
• Impact analysis: given a file, list all consumers to assess risk before edits