semantic-code-search

What problem does it solve?

Traditional keyword searches (grep/ripgrep) are inefficient for large codebases, leading to too many false positives and high token consumption. This Skill provides semantic understanding to find code by its function, not just its name.

Core Features & Use Cases

• Semantic Understanding: Find code by what it does (e.g., "user authentication login flow") across millions of lines.
• Token Optimization: Reduce AI token usage by ~40% by retrieving only relevant snippets.
• Hybrid Search: Combines keyword matching (BM25) with dense vector embeddings for superior relevance.
• Use Case: Quickly locate all instances of "payment processing with Stripe integration" in a large, unfamiliar codebase without sifting through hundreds of irrelevant keyword matches.

Quick Start

Use the semantic-code-search skill to index the current project directory. Then, search for "user authentication login flow with JWT tokens".