Comment by chongli

What if instead of allowing the agent to act directly, it writes a simple high-level recipe or script that you can accept (and run) or reject? It should be very high level and declarative, but with the ability to drill down on each of the steps to see what's going on under the covers?