Though I find that you shouldn’t “rely” on GPT for security guidance, I do believe you can use it more effectively. Like everyone already wrote, the author shou

Though I find that you shouldn’t “rely” on GPT for security guidance, I do believe you can use it more effectively. Like everyone already wrote, the author should have also queried GPT about the security implications of code execution if that’s the goal of the authors evaluation. This was an example of bad prompts, not leading to a (hidden) desired goal.