@S1r1u5_ It's definitely wild time to the whole internet
The movies about hackers which hacks literally anything is not just movie magic it's reality now
@CristiVlad25@AnthropicAI@bcherny I guess, one way to figure it out is to granulate tasks for even small pieces to make it harder for Claude to fool you.
It's better to feed ai with small pieces one by one instead of big bunches of them.
Claude is lazy or it's affected by context rotting with such effects
@0xtavian I thought the axiom works in a similar way. Thanks for answering. Now I'm using fleex for this kind of tasks. Will try your tool, thanks for sharing!
@0xw2w@damian_89_ I fucking hate that shit. It always looks like we are playing the stupid game "find the bug only on our specified endpoints" that doesn't have any relation to the company's real security.
CVE-2024-30043: @chudyPB details this #SharePoint XXE he discovered. He calls it one of the craziest XXEs he has ever seen, both in terms of vuln discovery and the method of triggering. He shows how it can be used for info disclosure & NTLM relaying. https://t.co/BzUDEE5Cy8