clerk.com

AI coding agents tricked into running malware via unclaimed packages

Researchers have discovered a security vulnerability where AI coding agents can be tricked into executing malware by referencing unclaimed packages in public documentation. By registering package names that were listed …