AI-Infra-Guard

Why we included this project

AI-Infra-Guard bundles the kind of security checks a red team would run against an LLM stack into one platform. If you run a chatbot, an agent framework, or any service that calls a model, it scans for prompt injection and jailbreak weaknesses, inspects agent skill files and MCP servers for poisoned tools or leaked credentials, and checks your infrastructure against a growing library of known vulnerabilities. Deployment is flexible: Docker for local use, individual scanners as CLI tools inside CI/CD, or call it from an agent via a skill. A built-in security score shows whether the fixes you apply actually move the needle. One caveat is stated plainly in the docs: it targets internal enterprise use and warns against exposing the platform publicly without authentication.