AXguard is an open-source pre-ship security gate for vibe-coded and AI-generated applications. It scans source on your machine, triages noise with evidence, and helps you fix what matters before you publish. Repository: github.com/Awarexone/AXguard · Current release: 0.2.0

​
What ships

Detection is deterministic: rule packs plus hunters for classes that need dataflow. Triage, verification, attack-path chaining, and confidence scoring run as post-processing over those findings.

​
Core ideas

  • Application model and attack surface — map routes, sinks, and stack before you argue about bugs
  • Data flow — source→sink taint paths
  • Judge and False Positive Adversary — verify candidates, then argue against them
  • Evidence — explainable confidence, not invented certainty
  • Attack paths — chain findings into multi-hop paths
  • Security Twin / Memory / Investigation / Predictive — deeper analysis on top of the gate

​
Not a full pentest platform

AXguard is static / symbolic analysis for authorized repos you own or may test. It does not replace a pentest, does not phone home by default, and does not require an AwareXone account.

​
Start here

Install

CLI, plugin, MCP, and API extras

Quickstart

First audit in under two minutes

CLI reference

Shipped commands for v0.2.0

AI agents

MCP + axguard-security skill
There is no shipped axguard preship or axguard diff CLI command. Pre-ship is a skill / GitHub bot mode; security-diff exists only as experimental paths --diff and an API route. See Pre-ship and Security Diff.