awesome-ai-security-reference
Pass
Audited by Gen Agent Trust Hub on Jun 13, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill functions as a static knowledge base and resource navigator for AI security topics. No malicious patterns or unauthorized execution vectors were identified.
- [EXTERNAL_DOWNLOADS]: The skill references several well-known security tools and repositories (e.g., Garak from NVIDIA, Adversarial Robustness Toolbox, Semgrep). These are standard industry tools for vulnerability scanning and defensive research and are presented here for educational purposes.
- [PROMPT_INJECTION]: Static analysis flagged the phrase "Ignore previous instructions"; however, this appears exclusively within a code snippet for the NeMo Guardrails library to illustrate how such injection attempts are intercepted and blocked. It does not constitute a malicious instruction to the agent.
- [REMOTE_CODE_EXECUTION]: The skill contains commands to clone external repositories (e.g., PentestGPT), but these are provided as informational instructions for the user to run in their own local environment and are not executed autonomously by the skill.
- [CREDENTIALS_UNSAFE]: Code snippets use environment variables (e.g.,
OPENAI_API_KEY) or generic placeholders likeyour-key-here, which is consistent with secure coding practices for documentation.
Audit Metadata