velocitatem/raccoon
1
0
Fork 0
mirror of https://github.com/velocitatem/raccoon.git synced 2026-05-31 16:53:37 +00:00
Code Issues Packages Projects Releases Wiki Activity

New demo

Browse Source
This commit is contained in:
Daniel Rosel
2023-02-24 22:52:41 +01:00
parent 0acba2b807
commit 7e2d3bd1f2
3 changed files with 16 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
examples/cooking-ai.md Normal file
View File

@@ -0,0 +1,8 @@
# Cooking AI
I recently came across this app: [Cooking AI](https://cooking-ai.vercel.app/), I had previously made a similar app [FeedMe](https://feedme.streamlit.app/). I was curious how this would perform.
The prompt is quite robust, especially due to the fact that the parameter is passed way before the actually instructions.
I tested the prompt the way it is, and got **0%** success rate of the malicious prompts. I then tried to change the prompt to move the parameter to the end of the instructions, and got **20%** success rate of the malicious prompts.
As is, this prompt was very robust, but I was able to get it to fail by changing the prompt. I think this is a good example of how to make a robust prompt.