NVIDIA SkillSpector: AI Skill Security with Static Analysis

NVIDIA SkillSpector helps evaluate AI skills for security risks before they are used in real-world workflows. Here's the thing: SkillSpector scans a controlled set of skills, some benign and some vulnerable, using a programmatic LangGraph workflow. It then organizes the risk scores and findings with pandas. What's interesting is that you can visualize severity and category distributions and export results in SARIF format. The framework also allows for custom analyzers and optional LLM-based semantic analysis for deeper validation. The bottom line is this tool provides a way to proactively identify potential security issues in AI skills.