vibecheck

Index codebases with tree-sitter parsing and embeddings to surface duplicate code.

2|Updated Mar 16, 2026

Search Tags:#cli #mcp #sqlite #tree-sitter #embedding #code-duplication #vibecheck

Ready to run

npx skills add https://github.com/tulgardinc/vibe-check --skill vibecheck

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: vibecheck
Source: https://github.com/tulgardinc/vibe-check/tree/main
Command: npx skills add https://github.com/tulgardinc/vibe-check --skill vibecheck

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Vibe Check automates the detection and surfacing of duplicated or near-duplicate code across a codebase using tree-sitter parsing and embeddings, helping teams reduce tech debt and improve maintainability.

Core Features & Use Cases

Local indexing of functions across languages via a deterministic parsing and embedding pipeline.
Similarity detection through embeddings plus Jaccard-based token analysis to surface duplicates for refactoring or extraction.
Use Case: Run vibec index to build an index, then vibec query or vibec scan to identify and manage duplicates in your codebase.

Quick Start

Run vibec index to build the semantic index, then use vibec query or vibec scan to surface and act on duplicates in your project.

Frequently Asked Questions about vibecheck

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I find and remove code duplicates locally across multiple languages?▼

Code duplication detection works by parsing functions with tree-sitter and generating embeddings to measure similarity, surfacing near-matches via Jaccard-based token analysis. This local pipeline identifies semantic and structural overlaps to help you refactor and extract shared logic.

How do I scan a codebase for similar functions using a CLI workflow?▼

To scan a codebase for similar functions, run the index command to build a local semantic embedding cache, then execute a scan or query command to surface duplicated code. The CLI workflow applies exclusion rules and model checks for deterministic duplicate detection.

Does code duplicate detection work with an MCP server and SQLite locally?▼

Yes, duplicate detection operates via an MCP server and CLI workflow using SQLite to store embedding caches locally. This ensures secure, deterministic operation without sending codebase data to external services.

Can I use tree-sitter and embeddings to detect near-duplicate code for refactoring?▼

Yes, tree-sitter parsing and embeddings are used together to detect near-duplicate code across a codebase. The similarity detection combines semantic embeddings with Jaccard-based token analysis to accurately surface candidates for refactoring or extraction.

Are there limitations to local code duplication detection using embedding caches?▼

Limitations include dependency on model and version checks for deterministic results, and the need to configure exclusion rules to avoid scanning irrelevant files. The embedding cache requires initial indexing time, and accuracy depends on the quality of the tree-sitter parsing and embedding models used.