ClawKeeper

Why we included this project

ClawKeeper is a security framework for OpenClaw agents, built to catch the actions an agent takes that you never approved. It protects through three layers that sit at different points in the pipeline: skills inject safety policies into the agent's context, a plugin hardens configuration and watches behavior during execution, and a watcher runs outside the agent and can halt high-risk actions or demand human confirmation. The watcher layer is the most distinctive: because it is decoupled from the agent's internals, it works across different agent platforms and can be deployed locally or in the cloud. For teams running agents in enterprise or intranet settings, that separation between task execution and safety enforcement is a concrete pattern to study and adapt.