After configuring a Skill in WorkBuddy, my daily runs still come back empty—did I write something wrong?

Title: How detailed do classification rules in WorkBuddy's Skill actually need to be? More than a hundred messages all went into "Other"

Teaching WorkBuddy to do the work took more effort than doing it myself.

My shop gets over a hundred customer service messages a day. I've long wanted to split them into four categories—refunds, exchanges, shipping reminders, and size questions—so I can glance at them in the morning and be done. I used Excel for three months before, and I was pretty smooth with the table-merging routine, so this time I wanted to try Skills. The tutorial said once you set it up, it runs automatically every day.

I wrote three paragraphs in the Skill. The first described the role, the second listed the four categories, and the third said the output should have four columns: time, customer nickname, category, and original text. After saving, I hooked it into a scheduled task and set the time for 7 a.m.

Yesterday at 7:30 I turned on the computer. There was indeed a file sitting in the folder. I opened it and there was just a header row, with everything below blank. The status showed success—I'll look into that later. What confused me even more was the manual run. I dragged that day's chat logs in and ran it. This time it did something, but the category column had more than a hundred rows all marked "Other." The txt I exported has a timestamp at the start of every line. I suspect it can't recognize that, so I deleted the timestamps and tried again—still all "Other." I clearly wrote the four category names: refunds, exchanges, shipping reminders, size questions, not a single character off.

What I most want to figure out right now is how detailed the classification rules in a Skill actually need to be. Do I just give it the four category names, or does each category need two or three examples—like under refunds, writing actual phrases such as "I want my money back" or "Can this be refunded?" Also, shipping reminders and size questions sometimes appear in the same sentence. How should priority be ordered? Do I need to spell that out in the rules too, or else does it just dump everything into "Other"?

Has anyone used it this way? Please give me some pointers. I've spent almost three hours on this over the past two days, and the time I saved hasn't caught up with what I spent.