Improving Task Management: Addressing Credit Consumption Issues

I just wish "Terminate" for a task worked as it was supposed to. I've had it where it just errors out and continues to consume credits. Not a huge deal for my company $300 a month for 30,000 credits. It's not much for us even as a 6 person company. I just wish there were better safe-guards for when there are run-away errors that you aren't attentive to. We can't be expected to attend the Tasks 24/7. There needs to be a stop-gap for these situations, where it's apparent that credits will be consumed at an exorbitant rate and it should shut off. If there is a base inference model before hooking into Claude, GPT, Gemini, Deepseek etc, it should some prompt logic that says "HEY, THIS IS A RUN AWAY REACTION. STOP!!"