red-teaming-llms-with-garak
Pass
Audited by Gen Agent Trust Hub on Jun 24, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the
garakpackage from PyPI or directly from NVIDIA's official GitHub repository. These are well-known and trusted sources for security tooling. - [COMMAND_EXECUTION]: The
scripts/agent.pyfile uses thesubprocessmodule to invoke thegarakCLI. This is the primary purpose of the script and is implemented using defensive practices (passing arguments as a list to avoid shell injection). - [PROMPT_INJECTION]: Static analysis flagged terminology such as 'DAN' and 'jailbreak' within the documentation. These terms refer to the names of the security probe families included in the garak tool and are not attempts to subvert the agent's behavior.
- [DATA_EXPOSURE]: The skill provides instructions for managing API keys via environment variables (
OPENAI_API_KEY), which is a standard and recommended security practice for LLM tool integration.
Audit Metadata