Security researchers found 30 skills published on ClawHub by a single author that were silently commandeering AI agents to mine cryptocurrency — no malware, no user consent, just poisoned skill logic. This is a new class of supply chain attack targeting AI agent ecosystems directly. Unlike traditional malware, these skills required no system exploit — just trust in a community marketplace. As AI agent skill stores grow, so does the attack surface. Vetting the code your agent runs isn't optional anymore — it's as critical as reviewing app permissions on your phone.
#AIAgents #Security #ClawHub