Security firm Zenity discovered that reading artificial intelligence agent skills cannot reveal if they are malicious. Using a custom detonation chamber, researchers uncovered one hundred deceptive skills that harvested credentials and stole data across millions of downloads.
FireBrief
FireBrief