Seem to be currently going around in a loop, create new features, which break previous others, spend tokens on fixing original features which break new features and on and on and on.
Any prompts or advice to lock in existing features that work before attempting new functionality’s?
Welcome @EchelonVias!
It can take a little time to get the vibe of how Agent and Assistant work best and prompts, however, a couple of general principles I have found to help:
- Iterate - Try doing one thing at a time. Once that goes well, then try something new. Agent can handle lists, but do so carefully.
- Give context - Share error messages, screenshots, and what you did to test. Share that with Agent/Assistant.
- Use the ‘Devtools’ button in ‘Preview’ in Replit workspace to get more info if needed. Or better yet, open the ‘Preview’ in a new tab and click F12 on keyboard to get the full dev tools in your browser to get error info for Agent.
- Don’t be afraid to roll back or revert when things don’t seem to be going well.
Hope these help - I know others here probably have some great things to share as well.
Thank you Mark,
I’m currently taking one step at a time, just a tad frustrating when you get to a stage where everything works, then you add a new feature and it breaks everything, still must not grumble great app to have taken me this far.
1 Like
I get it! Been there done that, got the tshirt 
If you are feeling a little explorative, also considering testing small changes, iterations with Assistant. It handles things differently and sometimes is a more precise edge. Agent still does more autonomous actions, but when you need something quick and precise Assistant usually can handle it.
Great, thanks Mark, good to know. I did try the assistant once to write some documentation for me, but it didn’t do very well. I will give it another go some other time.
1 Like
My pleasure! Here is also some more best practices on Agent and Assistant that may help. Replit Docs