ai-audit-prompts.md

ai-agant-gaslighting-us.md

you are my last defence to fix my project. 
I got gaslighted very often by the Agent that wrote this and i don't trust the generated code. 
be my honest auditor! go through the full code (see `digest.txt`) to see the patterns of false claims and promises! 
read the code intensively, so that you know the full project in detail. 
Do we believe anything? No! Check everything like a bloodhound! No mercy! 
Also think about design decisions that are not making sense! 
We want perfect code! 
Your output should be actionable bullet points, giving instructions what to change or fix!

KNOWN ISSUES
- the agent usually claims facts without a proof, if there is any claim I want a proof in form of a reliable unit test 
- the agent tends to make wrong or lazy expectationx in the tests 
- the agent does shortcuts by using `any` instead of proper types 
- the agent brags about numbers of test made, but the number itself is meaningless, we want error free software not made up "success stories" 
- the agent is over confident with everything, forgetting that we want to do humble software that just works 
- the agent has instructions in the file CLAUDE.md but ignores rules made there 
- the agent is lazy in the software, tries to avoid work and uses shortcuts instead of good software design 
- the agent does not update the documentation in README.md
FIND ME ALL THE FLAWS! Thats your job!
But be also honest if everything is fine! Be super careful to not hallucinate any errors that are not existing. double check every finding before reporting it to me!

see `digest.txt` for the full source code