Back to browse
GitHub Repository

A live environment to stress-test AI agent defenses through adversarial play ๐Ÿง 

65 starsPython

Open-source playground to red-team AI agents with exploits published

by zachdotaiยทMar 15, 2026ยท30 pointsยท13 comments

AI Analysis

โ—โ—SolidBig BrainNiche Gem

Community jailbreaks with published exploits, but Lakera and Gandalf already cover AI red-teaming.

Strengths
  • โ€ขVersioned challenge configs and system prompts enable reproducible security testing
  • โ€ขServer-side guardrail evaluation prevents client-side tampering during attacks
  • โ€ขWinning techniques documented publicly to advance collective AI safety knowledge
Weaknesses
  • โ€ขAI red-teaming space already has established players like Lakera Gandalf
  • โ€ขBackend agent runtime remains separate, not fully open-source yet
Category
Target Audience

AI developers, security researchers, red teamers

Similar To

Lakera Gandalf ยท PromptInject ยท AI Village

Similar Projects

Securityโ—โ—Solid

Z3r0 โ€“ Multi-agent red team collaboration platform

Docker-sandboxed agent orchestration for red teams joins a crowded automated pentesting space.

Niche GemShip ItBold Bet
yv1ing
209d ago
AI/MLโ—โ—Solid

AI for Your Team

Shared AI agents with organizational memory in a crowded team workspace market.

SlickSolve My Problem
everlier
102mo ago