Skip to content

anthril/vibe-guard

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

64 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

VGuard

AI coding guardrails that actually enforce.

npm version npm downloads license CI node


AI coding tools write code fast, but they also introduce security vulnerabilities, break project conventions, and push to protected branches. VGuard sits between the AI tool and your codebase, checking every proposed change before it happens. Bad changes get blocked with a clear explanation and a suggested fix. Good changes pass through without friction.

VGuard architecture flow: Developer Prompt → AI Agent → VGuard → Pass or Block

Quick Start

npm install -D @anthril/vguard
npx vguard init

Answer four questions and you have working guardrails. The init wizard detects your framework, asks which AI tools you use, and generates everything. See the full Getting Started guide.

Features

Runtime Enforcement — Claude Code hooks run before every tool call. VGuard inspects the proposed change, evaluates it against your rules, and blocks anything that violates them. The AI agent sees exactly what went wrong and how to fix it. Adapter docs →

Advisory Guidance — Cursor, Codex, and OpenCode don't support runtime hooks, so VGuard generates configuration files that teach the AI your project's rules before it starts writing. Agent setup →

Smart Detection — Edit rules only flag problems that are newly introduced. Pre-existing issues in the file are left alone so you can adopt guardrails incrementally without fixing every legacy violation first. Rules overview →

Rule Layers

Five rule layers: Security, Quality, Workflow, Custom, Analytics

Rules — 25 built-in

Category Count Examples Docs
Security 7 branch-protection, secret-detection, prompt-injection, rls-required security rules
Quality 11 import-aliases, naming-conventions, hallucination-guard, dead-exports quality rules
Workflow 7 commit-conventions, pr-reminder, migration-safety, changelog-reminder workflow rules

Presets — 14 ecosystem presets

nextjs-15 · react-19 · typescript-strict · supabase · tailwind · django · fastapi · laravel · wordpress · react-native · astro · sveltekit · python-strict · go

Browse all presets →

Agent Support

Agent support: Claude Code (runtime), Cursor/Codex/OpenCode (advisory), GitHub Actions (CI)

Documentation

Section Description
Getting Started Install, init wizard, first run
Configuration vguard.config.ts reference
CLI init, add, remove, generate, doctor
Rules All built-in rules with examples
Presets Framework-specific rule bundles
Agent Setup Per-agent adapter configuration
Custom Rules Write your own guardrails
Troubleshooting Common issues and fixes

VGuard Cloud

VGuard Cloud gives your team a dashboard for AI coding activity — which rules fire most, who's triggering blocks, and how conventions drift over time. Set up drift alerts, connect webhooks, and export analytics. Free tier available for small teams.

Cloud docs → · Sign up →

Sponsors

This project is maintained by Anthril and funded by our sponsors.

Become a sponsor →

Featured Sponsors

All Sponsors

Community

License

Apache 2.0 — Anthril

About

AI coding guardrails framework. Runtime-enforced quality controls for Claude Code, Cursor, Codex, and more.

Topics

Resources

License

Contributing

Security policy

Stars

Watchers

Forks

Sponsor this project

 

Contributors

Languages