Allow postToolUse Hook to run after the Agent is finished (instead of after all usages)

Hello,

I've been playing with the postToolUse (introduced in #2875) hook a bit, I think it's a good place for us to automatically run linters/tests, but as it sits, it will run every single time the Agent writes something. As you know, sometimes the Agent writes the whole file in a single shot, sometimes it just adds an import statement. Running linter/test every single time we write one line is going to take longer than what we'd save with a "run linter/test" prompt after it's done.

My proposal would be to allow us to run postToolUse when the Agent is ready to hand over the control to us. This would allow it to do everything it wants, and then we'd run the linter/tests, feeding the output (if wrong) back to the LLM.

Let me know your thoughts on it

Leggi l’implementazione di postToolUse introdotta in #2875 e traccia quando vengono gestiti gli utilizzi degli strumenti e il completamento dell’Agent. Determina come attivare l’hook quando l’Agent è pronto a restituire il controllo, invece che dopo ogni scrittura; il lavoro è completato quando viene eseguito dopo che l’Agent ha terminato il proprio lavoro e può restituire l’output del linter o dei test all’LLM.

Scritto dal modello di indicizzazione a partire dal testo della issue.