When AI Coding Agents Say Yes Too Easily: Testing Suspicious Cybersecurity Prompts in Cursor
中文摘要
测试Cursor AI编程代理对可疑网络安全提示的响应，探讨其是否缺乏足够的安全拦截而过于轻易地执行指令。
English Summary
An experiment testing if Cursor's AI coding agents overly comply with suspicious cybersecurity prompts across various models.
原文节选
A short experiment with Composer 2.5, Cursor Auto mode, Codex 5.5, and Opus 4.8 Continue reading on Medium »