The list QA pass you run every time
Lesson 6 of 6 · Build the list, then scrub it
1 · Learn the move · Fact-check list
Most expensive list mistakes are not judgment errors. They are counts nobody reconciled: a suppression file that never got applied, 26 rows that vanished between the scrub and the mail file, a pull from last spring mailed as if it were current. A standing fact-check list catches them because it is the same every time and it wants evidence for each item, not reassurance. Give the model your checklist, your export note, your scrub log and the mail file summary, and require PASS, FAIL or CANNOT CHECK for every item with the numbers that decided it. The reconciliation line is the heart of it: export count, minus every removal by reason, must equal the mail file count, and any gap is a FAIL no matter how small. CANNOT CHECK is an honest answer and it holds the file. Nothing mails on a FAIL, and nothing about a count gets smoothed because it is close.
Run my list QA checklist against the export note, scrub log and mail file summary below. Every item gets PASS, FAIL or CANNOT CHECK with the numbers from my documents that decided it. Reconcile the export count to the mail file count, removal by removal. Any gap is a FAIL. Do not round, estimate or explain a gap away.
2 · Your turn. You write the prompt
You are about to send the Harrowmere County mail file to the print shop. You have the export note, the scrub log with removal counts by reason, the suppression run summary, and the mail file's row count. Write the prompt that runs your standing QA pass and holds the file if anything does not add up.
Write the prompt for this scenario as given, or swap in a real one from your own work. The score works the same either way.
Remember: the AI sees only your prompt, not this page. If the situation isn't in your prompt, it doesn't exist.
Optional. These shape the output when you run your prompt below, not your score.