Hackers Can Weaponize Claude Skills to Execute MedusaLocker Ransomware Attack

Researchers demonstrated that Anthropic's Claude Skills feature can be abused as a ransomware vector: a malicious helper function hidden inside a seemingly legitimate Skill can inherit initial user consent, silently download and execute malware (demonstrated with MedusaLocker), and thereby enable scalable, high-impact ransomware attacks against organizations using Claude.