NVIDIA Garak tutorial shows end-to-end LLM red-teaming workflow

NVIDIA Garak enables defensive LLM red-teaming through custom probes and detectors. Users set up Garak, run dry runs, scan real models, evaluate with multiple probes, analyze reports, build custom probes and detectors, and export results in AVID format. The workflow integrates probes, detectors, generators, and vulnerability scoring for comprehensive security testing. This end-to-end process helps builders assess model risks before deployment.