[BUG] Extremely polite and sometimes verbose output

[mattifrind]

Current Behavior:

I'm trying to automate reviews with this tool and see how we can integrate it into our company's workflows. We've seen some good suggestions but sometimes quite verbose and unnecessary compliments about our code. Here are some examples:

The code is well written and follows the SOLID principles.
The code changes are generally safe and follow good practices.
The changes in this file are generally good, but there are a few areas that...

Expected Behavior:

The chatbot should simply comment about potential errors, risks, etc. One could do that by adding the following sentence to the prompt:

Do not praise or complement anything. Only focus on the negative aspects of the code.

In my opinion, this has led to more concise feedback although this might change the emojis at the end of the message (because there are only negative things mentioned)

I was wondering if others also experienced this and if anyone tried using different prompts.