A security vulnerability exists in six major coding agents: via a repository with malicious code, attackers can trick the AI ...