Do you feel confident when coding agents ask for permissions?

That part of the workflow still feels too loose. Agents are getting better at writing code, but permission requests and decision-making are often opaque, inconsistent, and hard to audit.

If an agent is going to act on my behalf, I want clear prompts, a clear scope, and a clear trail of what was approved and why.

The real question is not just “can the agent do the task?” It’s “Can I trust how it asks to do it?”

That's why I thought of working on a tool as a side project called Fridge.

I'm going to open the private beta soon. You can join the waitlist at: https://fridge.sh