Skill Detail
Scan LLM systems for jailbreaks, prompt injections, and unsafe behaviors with garak
Probe a model or agent stack with adversarial test suites so safety failures show up before deployment or review.
Security & VerificationMulti-Framework
Security & Verification
Multi-Framework
Security Reviewed
⭐ 7.5k GitHub stars
COPY SKILL INSTRUCTIONS (OPTIONAL)
npx skills add agentskillexchange/skills --skill scan-llm-systems-for-jailbreaks-prompt-injections-and-unsafe-behaviors-with-garak
Uses the third-party skills CLI, not an ASE-owned installer. Check your agent’s compatibility. This copies instructions; complete the upstream tool setup below separately.
At a glance
Tools required
Python 3.10+, target LLM or API credentials, command line access
Install & setup
Install with
python -m pip install -U garak, configure access to the target model or provider, then run garak with the generator and probe options that match the system you want to assess.Author
NVIDIA
Publisher
Organization
Last updated
Apr 16, 2026
Quick brief
Use garak when the operator needs an active vulnerability scan of an LLM system, not a passive policy checklist. It runs named probe suites against target models and reports jailbreak, prompt injection, leakage, toxicity, and related failure modes across many providers. The boundary is red-team style model scanning and assessment, not a generic security platform or a broad repo scanner.