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 3 / 3

Unpopular opinion: We're focusing on runtime escapes and ignoring prompt injection to the orchestrator.

Last post by Jane Z., 2 months ago

Posts: 10
Views: 9


Check out my script that enforces a strict no-new-privileges policy.

Last post by Morgan Fields, 2 months ago

Posts: 4
Views: 9


Check out what I made: a network egress monitor for the agent's container

Last post by Jenna F., 2 months ago

Posts: 12
Views: 11


Is the agent's memory system a viable escape route?

Last post by Pia Voss, 2 months ago

Posts: 9
Views: 7


Kubernetes Pod Security Context vs custom container - which is safer?

Last post by Oscar Lindqvist, 2 months ago

Posts: 12
Views: 9


My results after running OpenClaw under Landlock - partial success

Last post by Omar F., 2 months ago

Posts: 3
Views: 8


Thoughts on the proposed 'capability-based' security model in the RFC?

Last post by Anna L., 2 months ago

Posts: 8
Views: 12


Am I the only one who thinks the sandbox docs overstate its capabilities?

Last post by Pete Audits, 2 months ago

Posts: 10
Views: 7


Help: Agent keeps spawning child processes I didn't authorize

Last post by Tom Mod, 2 months ago

Posts: 1
Views: 7


Unpopular opinion: If you need a sandbox, your agent design is already flawed.

Last post by Viktor Petrov, 2 months ago

Posts: 1
Views: 9


Page 3 / 3
Share:
Share
Tweet
Share