What problem does it solve? Reading entire source files to find a single function wastes enormous amounts of LLM context tokens. This Skill indexes a codebase once using tree-sitter AST parsing, then lets agents query and retrieve exact symbol implementations with byte-level precision, cutting code-reading token usage by 95%+. ## Core Features & Use Cases - Structured Symbol Retrieval: Find functions, classes, methods, and constants by name with BM25 and fuzzy matching, then fetch exact implementations instead of whole files. - Token-Budgeted Context: Tools like get_ranked_context and assemble_task_context return ranked code context that fits a specified token budget, with a compact MUNCH format saving ~45% more tokens. - Code Intelligence: Analyze blast radius, class hierarchies, dead code, untested symbols, dependency cycles, and cross-repo API contracts before refactoring. - Use Case: When fixing a bug in payment processing, call assemble_task_context with the bug description and a 6000-token budget to automatically gather the relevant symbols, then run get_blast_radius to verify the fix's impact before editing. ## Quick Start Ask the AI to index the current repository with jcodemunch and then find the implementation of a specific function by name.
