Disclosed weakness · no CVE recorded
OpenClaw could start a writable sandbox without protecting its managed skills
OpenClaw · OpenClaw · sandbox first launch
First documented disclosure · Review checkpoint
Summary
A new sandbox could start without the intended read-only protection when its managed skill directories had yet to be created.
Severity and exploitation evidence
No CVSS score recorded in this review.
Exploitation not established: The upstream sources document the behavior and correction. They identify no malicious use.
Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.
How the weakness works
Sandbox initialization omitted the guard for directories absent at launch.
Conditions required
A writable sandbox starts for the first time with a missing managed skill directory.
Potential impact
Untrusted work could write content where the operator expected read-only managed skills.
Affected versions and configurations
The upstream report identifies the affected behavior without specifying a complete version range.
Fix and mitigation status
The 2026.11.4 stable release includes this issue-specific correction.
Practical next steps
Deploy the correction and validate skill protection when provisioning fresh workspaces.
Operating lesson
Check enforcement on the first run as well as on established installations.
Evidence and limits
The June 18 report specifically covers missing first-launch directories.
Sources
- OpenClaw · [Security/Bug] Sandbox 'skills' directory lacks initial read-only protection if host directory is missing upon agent creation
Primary source ·
- OpenClaw · fix(sandbox): prepare protected skill mount sources on first launch
Primary source ·
- OpenClaw 2026.11.4 stable release notes
Primary source ·