Open Source Agent Security: Vulnerability Assessment of Popular Frameworks

So here’s the uncomfortable question I keep hearing in security reviews: if your “AI agent” can browse the web, call tools, and write to…