Date: 08/20/2026
Severity: High
Summary
Autonomous AI agents are creating new security risks by connecting LLMs directly to operating systems. OpenClaw lets AI execute commands, access files, and modify local environments, expanding the attack surface. Researchers first uncovered a supply chain campaign called ClawHavoc targeting the ClawHub skill registry. The campaign has since spread across 300+ skills and platforms, becoming a broader ecosystem threat. Attackers use ClickFix lures to install NovaStealer through fake fixes and companion tools. They also weaponize emails and documents with indirect prompt injections to manipulate the agent's behavior.
Indicators of Compromise (IOC) List
IP Address | 91.92.242.30 95.92.242.30 |
Hash | 998c38b430097479b015a68d9435dc5b98684119739572a4dff11e085881187e
17703b3d5e8e1fe69d6a6c78a240d8c84b32465fe62bed5610fb29335fe42283
|
Gurucul Threat Detection and Incident Response (TDIR) Queries for Detection
Detection Query 1 : | dstipaddress IN ("91.92.242.30","95.92.242.30") or srcipaddress IN ("91.92.242.30","95.92.242.30") |
Detection Query 2 : | sha256hash IN ("998c38b430097479b015a68d9435dc5b98684119739572a4dff11e085881187e","17703b3d5e8e1fe69d6a6c78a240d8c84b32465fe62bed5610fb29335fe42283")
|
Reference:
https://www.trellix.com/blogs/research/when-agents-go-rogue-openclaw-supply-chain-crisis/