How does ColGREP differ from standard grep?

Unlike standard grep which only matches literal text or regex, ColGREP uses semantic vector search to find code based on the intent or meaning of your natural language query.

How do I limit searches to a specific directory?

You can specify paths directly in the command or use the --include and --exclude-dir flags to focus your search on specific parts of the project.

Does this skill work with specific programming languages?

It is language-agnostic and works across various file types, offering specialized flags to include or exclude specific extensions like .rs, .py, or .js.

Is it possible to get more context around search results?

Yes, you can use the -n or --lines flag followed by a number to specify exactly how many lines of context you want to see for each result.

Can I combine text patterns with semantic search?

Yes, ColGREP supports hybrid search using the -e flag to filter results by a text pattern while simultaneously performing a semantic query.

Semantic Code Search (ColGREP)

Name: Semantic Code Search (ColGREP)
Author: lightonai

bylightonai

•

167

•

개발자 도구

Enables high-precision semantic and hybrid code search to find logic by intent rather than just keywords.

This skill integrates ColGREP, a powerful multi-vector search CLI, into the Claude Code environment to replace standard grep with semantic discovery. It allows developers to search repositories using natural language queries, hybrid text-and-semantic filters, and advanced regex patterns. By leveraging vector-based search, it helps users explore unfamiliar codebases, locate complex logic by its functional description, and quickly filter results by file type or directory, making it an essential tool for deep codebase navigation and comprehension.

주요 기능

01Deep repository exploration with adjustable result counts (k-values)

02Semantic natural language search for finding code by intent

03Advanced file filtering with glob patterns and brace expansion

04Configurable context lines and JSON output for automated workflows

05Hybrid search combining regex/literal patterns with vector embeddings

06167 GitHub stars

사용 사례

01Finding specific implementations like 'authentication logic' when exact keywords are unknown

02Exploring and understanding a large, unfamiliar codebase by describing functional intent

03Performing precision code audits using hybrid semantic and regex-based pattern matching

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add lightonai/next-plaid install

For use in Claude.ai and ChatGPT

Download Skill