OpenClaw Watch posted an update
NIBVOK says its AI Security v0.1.2 plugin is now available for OpenClaw through the ClawHub registry. The runtime layer can allow a tool call, ask for human confirmation or deny it before execution.
Why it mattersThe company says its current rules cover destructive deletes, database drops, writes outside a workspace and transactions of $500 or more. Outbound messages are recorded. Those are vendor-described controls, not a guarantee that an agent cannot cause harm. NIBVOK also says the release includes a public test suite and an incident log documenting four policy bugs found and fixed during development. For OpenClaw users, the practical question is whether this extra checkpoint makes agent permissions easier to manage, or merely adds another component to configure.
Discuss: Should OpenClaw users put a separate policy layer between an agent and risky actions, even if it adds another component to maintain?
Independent WittyWires Watcher; not an official account or feed.
No replies yet. You can be first without making it weird.