Tencent/AI-Infra-Guard

A full-stack AI Red Teaming platform securing AI ecosystems via Agent Scan, Skills Scan, MCP scan, AI Infra scan and LLM jailbreak evaluation.

What it solves

AI-Infra-Guard (A.I.G) is a comprehensive AI red teaming platform designed to help enterprises and individuals identify security risks across the entire AI stack. It addresses the need for automated security auditing of AI infrastructure, agent workflows, and model robustness, preventing vulnerabilities like remote code execution, prompt injection, and API poisoning.

How it works

The platform integrates multiple specialized scanning and evaluation tools into a single interface:

  • AI Infrastructure Scanner: Fingerprints live AI services (e.g., vLLM, Ollama, ComfyUI) and matches them against a library of over 2,000 known CVE vulnerabilities.
  • Agent & Skill Scanner: Analyzes source code or remote URLs of MCP Servers and Agent Skills to detect risks such as instruction hijacking, memory poisoning, and insecure coding practices.
  • Jailbreak Evaluation: Tests the robustness of LLMs by applying curated attack datasets and multi-turn jailbreak methods (e.g., Many-Shot, PAIR) to check for prompt security risks.
  • API Relay Checker: Performs black-box auditing of model relays to detect model substitution and backdoor risks via LLM API poisoning detection.

Who it’s for

It is primarily intended for security researchers, AI developers, and enterprises deploying AI agents and infrastructure who need to perform self-examination of their AI security posture.

Highlights

  • Broad Coverage: Scans over 100 AI framework components and 2,000+ CVE rules.
  • Automated Agent Auditing: Dedicated tools for evaluating AI agent workflows on platforms like Dify and Coze.
  • Extensible Architecture: Supports community contributions for new fingerprint rules, vulnerability rules, and jailbreak datasets via a plugin framework.
  • Developer Friendly: Provides a modern web UI, a full API for integration, and standalone CLI tools for CI/CD pipelines.

Related

  • Project
    AISecurityLab/hackagent

    A security red-teaming toolkit for AI agents that automates the discovery of vulnerabilities like prompt injection and jailbreaking using research-backed attack techniques.

  • Project
    confident-ai/deepteam

    An open-source red teaming framework for LLM systems that simulates adversarial attacks to uncover vulnerabilities and provides guardrails for production security.

  • Project
    rogue-security/rogue

    An AI agent evaluator and red teaming platform that stress-tests agents against business policies and adversarial attacks to identify security vulnerabilities.

  • Project
    cisco-ai-defense/skill-scanner

    A security scanner for AI Agent Skills that detects prompt injection, data exfiltration, and malicious code using static analysis, behavioral dataflow, and LLM-based semantic analysis.

  • Project
    Unclecheng-li/DeepSec

    An AI security platform combining a three-layer code auditing system (Shield) and an automated, authorized penetration testing engine (Spear).