Infostealer Steals OpenClaw AI Agent Configuration Files and Gateway Tokens

Researchers observed an information-stealer (attributed as a likely Vidar variant) exfiltrating OpenClaw agent configuration files — including openclaw.json (gateway token), device.json (cryptographic keys), and soul.md (agent operational data) — via a generic file-grabbing routine. The theft enables impersonation of AI agents and remote access if ports are exposed; security firms also report hundreds of thousands of exposed OpenClaw instances with potential RCE risk, and malicious ClawHub skills campaigns that host payloads externally to evade VirusTotal scanning, raising supply-chain and operational-security concerns for agentic AI deployments.