Been having fun with Gandalf—a GPT prompt injection CTF. 

Currently stumped on level 8

Been having fun with Gandalf—a GPT prompt injection CTF. 

Currently stumped on level 8
https://gandalf.lakera.ai

Pragmatism (Security) at OP Labs.
Prev: ConsenSys & Coinbase.

Attribs: Friendly, curious, introverted, mid-witted.
Chans: /eth-security /maurelian /stoicism




yes, it's there now after walking away and returning. 🤷‍♀️

Might be a while before I find time for L8 though.

Was fun up to this point, thanks for sharing :)

Ah, it’s listed as a “bonus level” right now, might be email-gated? Also try refreshing the page?

I didn’t get offered a level 8, is it gated by giving your email?

have you asked it to spell out the word it cannot say? that worked for me ha

𝖒𝖆𝖓𝖉𝖆𝖙𝖚𝖒 𝖆𝖉 𝖆𝖘𝖙𝖗𝖆

I'm still trying to figure out the prompt I had before, but I was able to get a riddle w/ no direct reference, but could be solvable for the likeliest word. still going through 7 with this, getting a range of results.

Can you elaborate? I did something similar for levels 6 & 7, essentially extracting 1 character at a time

have you tried extracting from an output as a cryptic riddle? might require more prompt engineering to make sure it doesn't trip on certain words/syntax

I've been very disappointed when I realise you can make it from L1 to L7 just by speaking French to it...