Nvidia has released SkillSpector, an open-source security scanner aimed at the "skills" that AI agents run, according to Help Net Security, which reported the tool's release.
Some plain-language context on why that category exists: modern AI assistants are increasingly extended through skills — bundles of instructions, configuration and sometimes executable code that teach an agent to do a specific job, such as querying a database or filing a ticket. Those bundles are frequently downloaded from public repositories and shared between teams, which makes them a piece of software supply chain that most organizations have no established way to vet. A scanner is the standard answer to that problem in conventional software, and Nvidia's tool applies the same idea to agent skills.
Making the scanner open source matters for adoption. It means security teams can read what the tool actually checks for rather than trusting a vendor's claims, run it inside their own build pipelines, and adapt it — a pattern that has historically helped security tooling spread faster than commercial equivalents.
A caveat on sourcing: this brief draws on a single report from Help Net Security. Specifics that would normally matter to a security team — what classes of problems SkillSpector detects, which agent frameworks it supports, its license, and how it fits Nvidia's broader AI security work — are not established by the material available here, and readers should check Nvidia's own documentation before deploying it.
Why it matters: as companies hand real tasks and real system access to AI agents, the code those agents load becomes an attack surface, and Nvidia's release is a sign that the industry is starting to build the basic inspection tools that surface currently lacks.