rogue-security/rogue

AI Agent Evaluator & Red Team Platform

What it solves

Rogue is a security and evaluation platform designed to stress-test AI agents before they are deployed. It prevents AI agents from violating business policies or falling victim to adversarial attacks that could lead to security vulnerabilities.

How it works

The platform uses a client-server architecture to probe AI agents via several protocols (A2A, MCP, or direct Python calls). It offers two primary modes of operation:

  1. Automatic Evaluation: Tests agents against predefined business policies and scenarios to ensure compliance and expected behavior.
  2. Red Teaming: Simulates adversarial attacks using over 75 vulnerabilities and 20 attack techniques (such as prompt injection and social engineering) to find security gaps.

Results are analyzed by a "judge LLM" and scored using a CVSS-based risk system to determine the severity of vulnerabilities.

Who it’s for

Developers and security auditors who need to validate the behavior of AI agents, perform penetration testing, and ensure compliance with frameworks like OWASP, MITRE ATLAS, and the EU AI Act.

Highlights

  • Comprehensive Red Teaming: Includes 75+ vulnerabilities across 12 categories and 20+ attack techniques.
  • Flexible Integration: Supports Google's Agent-to-Agent (A2A) protocol, Model Context Protocol (MCP), and direct Python entrypoints.
  • Compliance Mapping: Aligns with industry standards including NIST AI RMF, GDPR, and ISO/IEC 42001.
  • Multiple Interfaces: Provides a TUI for interactive testing, a CLI for CI/CD pipelines, and a standalone server.
  • Deterministic Testing: Supports random seeds for reproducible security scans.

Related

  • Project
    Tencent/AI-Infra-Guard

    An AI red teaming platform that provides automated security scanning for AI infrastructure, agent skills, and LLM jailbreak robustness.

  • 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
    Kritt-ai/open-kritt

    An open-source security research platform that orchestrates AI agents to find, validate, and rank vulnerabilities in codebases using focused, parallelized workflows.

  • Project
    devchilll/scope

    A production-ready governance framework for enterprise AI agents that implements safety guardrails, compliance-as-code, IAM, and human-in-the-loop escalation.

  • Project
    plurai-ai/intellagent

    A multi-agent framework that stress-tests conversational AI agents by simulating thousands of realistic edge-case interactions to uncover failure points and optimize performance.