OpenClaw: Agentic AI in the wild — Architecture, adoption and emerging security risks

## Executive Summary The report describes emergent, real-world security incidents around OpenClaw — an agentic local AI assistant — including widespread internet exposure of its gateway (21k+ instances), rapid protocol-aware exploitation of exposed gateways, a malicious VS Code extension that installs a remote-access implant, and malicious third-party "skills" used to harvest browser and wallet credentials; it provides observed attacker behaviors and practical mitigation guidance.