The paperless file that finds itself
Lesson 5 of 6 · Run it like a shop
1 · Learn the move · Template pattern
A scanned deed named scan_0405.pdf is a document you own and cannot find. Renaming by hand is the hour nobody spends, and a model can propose names from the scan's text in seconds, if the naming template is fixed first and the empty-slot rule is fixed with it. The template is the lesson: parcel ID, document type, document date, county, in that order, joined by underscores, with document types from a closed list. The failure mode is quiet. A model that cannot read the recording date on a faded stamp will take a date from somewhere else - the notary line, the file's metadata, the day you scanned it - and the file sorts into the wrong place forever. So every slot is filled from the document's own text or carries a visible MISSING token, the model says which date it used, and any document naming two parcels is flagged rather than filed under one.
Propose a filename for each scanned document below using this template: PARCELID_DOCTYPE_YYYY-MM-DD_COUNTY.pdf. DOCTYPE from this list only: [list]. Every slot comes from the document's own text; a slot you cannot read is MISSING, never borrowed from another date or another file. Flag any document that names more than one parcel.
2 · Your turn. You write the prompt
You have 38 scanned PDFs from last quarter's purchases in Bressler County and Quillan County, all still carrying the scanner's names, in one folder. You have run text recognition on them and have the text for each. Write the prompt that proposes a filename for every one without guessing a single date.
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.