garak

Open detection-rule standard for AI agent security threats — like Sigma, but for AI agents. Executable rules across 10 categories; merged into Microsoft AGT, Cisco AI Defense, MISP, OWASP, FINOS & SigmaHQ. MIT-licensed.

Vulntrex - View, Run & Compare Garak Scans in one sleek web interface.

Projet issu du codelab Devfest Nantes 2025 “La guerre des prompts” : atelier de 2h pour apprendre à pirater des IA et comment les protéger via des frameworks open source

Out-Of-Tree Llama Stack Eval Provider for Red Teaming LLM Systems with Garak

Runtime control, security, and governance for AI agents — decide what an agent may do at the moment it acts, revoke that authority instantly, and keep tamper-evident evidence of every decision.

Python + SQL toolkit for LLM threat intelligence — flags prompt injection, jailbreaks, and data exfiltration mapped to OWASP LLM Top 10 (2025) & MITRE ATLAS.

Red teaming AI Goat™ to explore OWASP Top 10 for LLM Applications (2025), prompt injection, RAG poisoning, MCP-assisted auditing, and AI security vulnerabilities.

Automated prompt injection vulnerability assessment of local LLMs using Garak

Production-grade LLM evaluation pipeline for RAG chatbot — DeepEval + RAGAS + Garak + CI/CD | Financial domain | 7 metrics | Adversarial testing

LLM security scanner for jailbreaks, prompt injection, and system prompt leakage — Garak probes plus a hand-built attack layer, severity classification with deduplication, and a CI gate that fails builds on critical findings.

Turn AI red-team tool output into standard, sealed, provenance-stamped findings. Feed them into the systems your team already uses.

Sovereign LLM adversarial red-team scanner. 100+ attack probes via Garak. Local Ollama only. Zero cloud API exposure.

Iterative AI purple-team harness: 5 specialist agents under a deterministic state machine, ATT&CK + ATLAS aware. Includes reproducible findings against GPT-4o (EchoLeak-style injection, MCP tool-description poisoning) and an upstream PR for garak detector calibration drift.

On-demand, hard-capped uncensored red-team reasoner — bring it up and tear it down on RunPod. Model-agnostic; authorized use only.

Just playing around with Garak AI for security testing of LLMs

An OWASP-LLM-Top-10-2025 red-team teaching lab

Intentionally vulnerable RAG service (FastAPI, OpenAI): indirect prompt injection via poisoned documents leaks a system-prompt secret. Automated and scored with PyRIT and Garak.

IA ofensiva sin censura, contenida y segura para su operador — red teaming ético asistido por IA sobre Kali.

Relentlessly scanning, probing, prodding LLM endpoints. The Inquisition always gets its answers.

Automated LLM Security Evaluation Framework