Test and lint driven feedback to ZedAI

[AriMKatz]

Check for existing issues

• Completed

Describe the feature

https://www.cursor.com/blog/shadow-workspace

it's all explained there. When generating code, especially if there's a higher level composer concept, it would be good to give some objective function that consists of tests that need to be passed (maybe it can write them in the background if not already), and generally access to all the context a human would have, like lint errors etc. Then you can spend cycles iterating on this, so you're getting more agentic behavior, but constrained.

If applicable, add mockups / screenshots to help present your vision of the feature

No response