An open agent-security benchmark, including the attacks we fail to catchgithub.com1 point·AndrewGS··0 commentsOpen articleSaveView on HN