[muchael]

Super cool! Please let me know how it goes. Since agents are so good at writing code, we think letting the agent rewrite/test the code on failure is better than just using a prompt at runtime