OS Command Injection in OpenClaw - CVE-2026-100559
Published: November 3, 2026
Vulnerability details
The vulnerability allows a remote user to execute additional operating system commands without the expected approval prompt.
The vulnerability exists due to improper handling of escaped newlines in OpenClaw's exec command parser when evaluating commands against an allowlist or approval policy. A remote user can supply lower-trust model input containing escaped newlines so that the parser authorizes one visible shell word sequence while execution includes an additional hidden command, bypassing the expected approval prompt.