VibeGuard

•
7
•

Scans AI-generated code for security vulnerabilities and provides real-time AI agent firewall protection.

VibeGuard is a powerful, offline-first security and privacy scanner specifically engineered for applications developed using AI coding assistants. It meticulously identifies a wide range of vulnerabilities, from leaked API keys and open databases to SQL and prompt injection flaws, and even uncapped AI agent loops. Operating entirely locally with zero telemetry, VibeGuard integrates seamlessly into development workflows, offering features like an agent action firewall to prevent data exfiltration, a two-layer AI firewall for robust prompt injection defense, and a local MITM proxy for polyglot network interception. It also provides comprehensive compliance mapping and supports integration with 15 AI clients and CI/CD pipelines, making it an essential tool for securing modern AI-driven development.

주요 기능

01Scans AI-generated code for over 764 security rules, including leaked keys, prompt injection, and uncapped agent loops across 18 languages.
02Provides a real-time AI agent action firewall to prevent secret and PII exfiltration before actions are executed.
03Features a two-layer AI firewall for comprehensive prompt injection defense using both regex matching and semantic classification.
04Includes a local MITM proxy for polyglot network interception, inspecting all outbound traffic for security risks.
05Offers one-command layered protection (`vibeguard auto`) with a daemon, pre-commit hooks, shell guard, and proxy for continuous security.
063 GitHub stars

사용 사례

01Securing AI-generated applications against common vulnerabilities and AI-specific risks like prompt injection and data exfiltration.
02Integrating automated security scans into CI/CD pipelines and pre-commit hooks for continuous vulnerability assessment.
03Monitoring and controlling AI agent behavior in real-time to prevent unauthorized actions and data leakage.