Hacker News Reader

It should be noted that this tool uses a prompt [1] which does include the whole file, or in case of non textual content, the metadata of the image/video/audio file [2].

[1] https://github.com/jjuliano/aifiles/blob/ef529fd6281eaf8d373...

[2] https://github.com/jjuliano/aifiles/blob/ef529fd6281eaf8d373...

View on HN

Output from the language model is also being injected into a script that is then executed: https://github.com/jjuliano/aifiles/blob/ef529fd6281eaf8d373...

He argued below that he is not vulnerable to indirect prompt injection attacks (https://github.com/greshake/llm-security), but I think he is wrong.

Reply on news.ycombinator.com