symbols

The Symbols skill is an advanced codebase navigation engine designed for large-scale software projects. It enables AI agents to query symbols—such as functions, components, hooks, services, and types—without loading entire files into the context window. By leveraging a pre-generated graph of symbols chunked by domain and architectural layer, this skill provides precise context that significantly improves token efficiency. It is primarily intended for developers and AI agents managing complex codebases where full-file context is redundant or prohibitively expensive.

• Intelligent Symbol Querying: Locate specific code artifacts by name, kind, path, or architectural tag instantly.
• Architectural Layer Awareness: Supports filtering by project-specific layers like Routers, Services, Repositories, or Schemas, allowing you to load only the code pertinent to your current task.
• Domain-Specific Chunking: Organizes large repositories into manageable domains (e.g., UI, Web, API, Shared) to optimize context window usage.
• Reduced Token Consumption: Achieves a 60-95% reduction in token usage compared to standard full-file loading, with additional backend-specific layer splitting providing up to 80% extra efficiency.
• Seamless Integration: Works with configuration-driven extraction scripts (TypeScript/Python) to maintain an up-to-date source of truth for the project structure.

Usage notes and practical tips:

• Always run the extraction and tagging scripts (extract_symbols_*.py, add_layer_tags.py) whenever significant structural changes occur in the repository.
• Use the symbol tools when you need to understand existing architectural patterns before implementing new features or when debugging cross-domain service interactions.
• This skill is strictly for read-only navigation and analysis; it does not replace the need for direct file reading (Read tool) when you require the actual implementation details of a specific function.
• Input expected is usually a query string targeting a symbol name or a path filter; output provides concise symbol signatures, line numbers, and docstring summaries.
• Constrain queries to specific architectural layers (e.g., only loading 'services') to maximize performance in large backend projects.