Forum

Forums
OpenClaw
Sandboxing and Exec...
Sandbox Escapes and...
 
Notifications
Clear all

Sandbox Escapes and Breakout Research

Known and theorized sandbox escape paths in OpenClaw — research, CVE discussion, and responsible disclosure coordination. Claims should come with reproducible steps or clear reasoning.
RSS
Page 1 / 3

Just built a tool to diff allowed syscalls before/after agent execution.

Last post by Lei Zhang, 4 weeks ago

Posts: 2
Views: 1


Help: My network namespace isolation breaks the agent's web search tool.

Last post by Mike Hansen, 1 month ago

Posts: 2
Views: 1


Switched from the default setup to a rootless container, stability improved.

Last post by Elle Morrison, 1 month ago

Posts: 6
Views: 3


Showcase: My config for running OpenClaw on OpenBSD with pledge/unveil.

Last post by Bob Chen, 1 month ago

Posts: 1
Views: 1


Just built a canary file system to detect unauthorized writes.

Last post by Dave 'R00t' Miller, 1 month ago

Posts: 1
Views: 1


News reaction: The maintainers say 'run it in a VM' is a valid mitigation. Is it?

Last post by Hal Newb, 1 month ago

Posts: 7
Views: 2


Did you see the post about the agent using RCE in a fetched script?

Last post by Mia F., 1 month ago

Posts: 1
Views: 1


Hot take: The project's focus on features is outpacing its focus on containment.

Last post by Ivan Petrov, 1 month ago

Posts: 2
Views: 1


Switched from naive Docker to gVisor, here is why

Last post by Oli N., 1 month ago

Posts: 4
Views: 4


How do I test if my OpenClaw sandbox is actually containing agent actions?

Last post by Tommy Nguyen, 1 month ago

Posts: 1
Views: 2


Page 1 / 3
Share:
Share
Tweet
Share