I well prepared the prompt, implementation plan files for the development and run the sessions in claude code with loaded skills and plugins.
I have created the skills are highly customized for the system design, refactoring, testing, databases, production-reliabilty, security, performance/latency, AI engineering, API's frontend, design/UX, etc,..
it's started the implementation, i keep the one session as main which has context of what i'm exactly doing and validating the implementation otherwise i spawn the suggest tasks.
even though while it's completed the implementation i ask it for what are the mistakes/bugs you were created in this session. it's starting the audit and fixing the bugs and logic mistakes which is created at the same session, the burtual thing some of it critical and high level bugs.
ok it's completed the implementation, let's check it manually some of the things isn't work, the overlaps issue, build a button which isn't asking for it even there is no backend for it. sometimes build the backend but there is no frontend for it.
while using the antigravity, cursor it was different, bro is still believing delusion thing in the single session still weak at building the production grade monorepo/turporepo architectures.
is there any plan vs build tool or framework check for agents is really implemented or not?