Improper access control in OpenClaw - #VU128783
Published: May 1, 2026
Vulnerability identifier: #VU128783
CSH Severity: Low
CVSSv4.0: CVSS:4.0/AV:N/AC:L/AT:N/PR:L/UI:N/VC:H/VI:N/VA:N/SC:N/SI:N/SA:N/E:U/U:Clear
CVE-ID: N/A
CWE-ID: CWE-284
Exploitation vector: Remote access
Exploit availability:
No public exploit available
Vulnerable software:
OpenClaw
OpenClaw
Software vendor:
OpenClaw
OpenClaw
Description
The vulnerability allows a remote user to disclose sensitive information from out-of-workspace mounted image paths.
The vulnerability exists due to improper access control in native prompt image ingestion functions detectAndLoadPromptImages / loadImageFromRef when processing prompt text that references mounted out-of-workspace image paths in sandboxed runs. A remote user can reference a mounted out-of-workspace image path in prompt text to disclose sensitive information from out-of-workspace mounted image paths.
Exploitation requires sandbox mode, tools.fs.workspaceOnly=true, a reachable out-of-workspace mount path from the sandbox, and a vision-capable model path active for native prompt image loading.
Remediation
Install security update from vendor's website.