Powers intelligent semantic code search and codebase discovery using LLM-enriched indexing and vector matching.
This skill integrates claudemem v0.2.0 to provide Claude with a deep, context-aware understanding of any codebase. By utilizing Tree-sitter for AST parsing and LLM-generated summaries of files and symbols, it moves beyond simple keyword matching to true semantic discovery. It allows developers and AI agents to find code based on purpose, behavior, and architectural role, offering specialized search modes like 'navigation' for exploration and 'fim' for completion to ensure the most relevant context is always available.
主な機能
01Specialized search presets for AI navigation, developer queries, and code completion
02Dual-matching architecture that searches both raw AST code chunks and semantic summaries
03LLM-Enriched indexing that summarizes file purposes, exports, and symbol behaviors
04177 GitHub stars
05Role-specific AI instructions tailored for architects, developers, testers, and debuggers
06Local vector storage via LanceDB for fast, privacy-focused semantic retrieval
ユースケース
01Architectural discovery to map dependencies and entry points in a new codebase
02Finding specific function implementations based on behavioral descriptions rather than names
03Providing AI agents with optimized navigation context to prevent excessive file reading
