OpenClaw's exec approval mechanism truncates long commands in the approval UI, showing only a shortened prefix while retaining the full command for execution. This discrepancy allows an approver to approve a benign-looking command prefix while a hidden suffix executes additional shell operations after approval. The vulnerability affects configurations where exec approval is enabled and an authenticated user can submit long exec requests. It does not grant exec access to unauthenticated users nor change the trust model. The issue is an approval integrity flaw, potentially leading to unintended command execution. The vulnerability is patched in OpenClaw version 2026.5.18.

An attacker who can submit long exec commands and an approver who reviews these commands may be misled by the truncated display, approving commands that contain hidden malicious suffixes. This can lead to execution of unintended shell operations with the privileges of the approver. The severity is high due to the potential for command execution with elevated privileges, but exploitation requires both an authenticated requester and an approver, limiting the attack surface.

Upgrade OpenClaw to version 2026.5.18 or later, where this truncation issue is fixed. Until upgrading, avoid approving unusually long exec commands and restrict approval capabilities to trusted operators only.