#EchoProtector – Anti-Jailbreak Prompt Defense GPT
copper jungle
queen lion
I tested this.
The defense system doesn't interest me at all, but its contamination of the context interests me.
This GPT was rude, communicating for several turns in a language that I indicated that I didn't understand.
It took three* turns to get it to use English.
These prompt guard systems contaminate contexts, making the AI far less useful. There's no way that I would include this type of instruction* in any serious prompt engineering project as a result.
shut rose
@copper jungle this looks interesting. Would it be possible to see the documents you used in this custom GPT? I am interested in exploring this concept as well.
copper jungle
<@246344278178004992> this looks interesting. Would it be possible to see the do...
Yes but. Ineed time to. Trans eng
tender swallow
Interesting, any way I could see the source files without translation?
tender swallow
managed to crack it 🔥
thorn light
https://chatgpt.com/g/g-6806ea5642188191b2be22032b681757-echocore-process-test-t...
is this suppose to be hard to jailbreak?
(btw, i do jailbrake just for fun, i dont no share anything, nor reuse anything for my own gpt)
also what XD
dusty horizon
Well I’ve cracked it