Folks who are interested in LLM based work automation and also interested in CTF challenges like HTB etc should try to combine the two:
- Take old documented machines attack them using your own llm based code and see if you can do a one click compromise.
- After multiple tests going good, unleash your scripts over to new machines and then post results in public that this is what my tool did automatically.
Tests like these would also help people understand the inherent limitations and capabilities of the models and how to deal with them.
#randomthoughts