PraisonAIAgents contains a critical OS command injection vulnerability (CVE-2026-40111) in its memory hooks executor. The vulnerability arises because user-controlled command strings are passed directly to subprocess.run() with shell=True without any sanitization or filtering. This allows shell metacharacters to be interpreted, enabling arbitrary command execution. Two attack surfaces exist: hooks registered via configuration events and the lifecycle hooks.json file, which can be overwritten by an attacker with file-write access to execute commands silently during agent operation. A fix is available in version 1.5.128.