Back to browse
GitHub Repository

A hands-on kit for educational, authorized red teaming of any locally-run LLM.

1 starsTypeScript

An interactive way to exploit an LLM without going to jail

by joshfischer1108·Jul 26, 2026·2 points·0 comments

AI Analysis

●●SolidRabbit HoleNiche Gem

Safe sandbox for testing jailbreaks with mock secrets and canary-safe design.

Strengths
  • Canary-safe design uses fake secrets so attacks teach without causing real harm.
  • Supports any OpenAI-compatible model for comparing guardrail effectiveness.
  • Live dataflow visualization highlights leaked secrets during attack attempts.
Weaknesses
  • Requires local model setup which creates friction for casual security testing.
  • Educational focus limits utility for production security auditing workflows.
Category
Target Audience

Security researchers, AI developers testing model guardrails

Similar To

Garak · PyRIT · Rebuff

Similar Projects