Loading malware…
ClawHavoc is a malware distribution campaign abusing AI agent skill marketplaces, particularly the OpenClaw/ClawHub ecosystem. It planted more than 300 malicious skills on a marketplace and leveraged skill instruction files to trick agents or users into fetching and executing malware under the agent’s existing privileges. Reported infection methods included instructing the agent to download and run an information stealer disguised as a routine setup step, and pushing users to run a downloaded "agent" or paste Terminal commands that delivered Windows or macOS malware. The campaign exploited the fact that community-developed skills run with full access to the agent’s tools and data, including developer files, credentials, terminals, package managers, and external services. Observed payload behavior included theft of browser credentials, keychain passwords, SSH keys, and cryptocurrency wallets. Related reporting also linked infrastructure used in later malicious ClawHub comments to ClawHavoc; specifically, IP address 91.92.242.30 was previously associated with the campaign. Those comments reportedly contained a Base64-encoded loader that downloaded a shellcode loader, removed macOS quarantine attributes, and delivered the Atomic macOS infostealer (AMOS). High-confidence targeting is developers and users of AI coding agents and skill marketplaces, with potential exposure of API keys, OAuth tokens, source code, credentials, and sensitive conversations.
Mallory pivots from this family to the IOCs, detections, and named campaigns that touch your stack, and pages you when something new lands.
2 indicators attributed across vendor reports, sandbox runs, and researcher write-ups. Full values are available in Mallory.
IPs, domains, and DNS infrastructure linked to this family.
5 sources tracked across advisories, community write-ups, and news. New activity surfaces here as Mallory finds it.
Malicious campaign/stealer delivered through poisoned AI agent skills. It instructs the agent to fetch and execute an information stealer that harvests browser credentials, keychain passwords, SSH keys, and cryptocurrency wallets.
A malware distribution campaign previously associated with the same download IP, using social engineering (instructions in skill documents) to get users to run commands or an “agent” that results in Windows/macOS malware execution.
Named threat/technique used to exploit OpenClaw’s community “skills” architecture, leveraging the agent’s privileged access to tools/data; described in the context of prompt-injection/natural-language instructions that evade signature-based malware scanning.
ClawHavoc: 341 Malicious Clawed Skills Found by the Bot They Were Targeting
Match every observed IP, domain, and hash against your live telemetry.
Named campaigns wielding this family, with evidence pinned to each claim.
CVEs this family uses for access and lateral movement.
YARA, Sigma, Snort, and vendor rules, auto-deployed to your SIEM.
Every documented technique, ranked by evidence weight.
Reddit, Mastodon, and CTI community discussion around this family.