Clawsec

Clawsec is an open-source security plugin for OpenClaw agents. It is built to add guardrails for agentic AI systems that can otherwise run with full system access, and it addresses risks such as prompt injection, secret leakage, unauthorized purchases, and data exfiltration.

Its middleware intercepts each tool invocation before execution. The page says that shell commands, HTTP requests, and file writes pass through three stages: interception, risk scoring, and action enforcement. The rules engine matches calls against built-in rulesets such as destructive-commands, secrets/, and exfiltration/, and the configured response can block the action, require human confirmation, or allow it with a full audit trail. The listed coverage includes destructive commands, secret exposure, data exfiltration, unauthorized purchases, privilege escalation, and prompt injection. The evaluation is described as completing in under 5 milliseconds.

Clawsec is delivered as an OpenClaw plugin, with installation shown as a single command: openclaw plugins install clawsec. The setup section says that it auto-generates a clawsec.yaml configuration file, and the configuration example shows extensible YAML rules. The page also describes it as open source, and states that it is MIT licensed. It presents itself as security for OpenClaw and as a tool for protecting AI agents.