Assemble a document-extraction prompt
Assemble a document-extraction prompt
Six supplier contracts sit in a folder as plain text. Finance needs renewal and termination details in a spreadsheet by Friday — and every value must be traceable to the sentence it came from.
Fill every field to assemble a prompt that extracts one row per contract — then compare with a strong example.
A strong example (self-assessment — not machine-graded)
Extract one row per supplier contract from the six texts below. Columns: Supplier | Term end | Notice period | Renewal type | Source quote. The Source quote column must contain the exact sentence each row's values come from. If a value cannot be quoted from the text, write 'not stated' in its cell instead. Copy dates and periods verbatim from the text. Do not calculate, convert or normalize any date or duration. Output only the table, with no commentary around it. Contracts: [paste the six contracts]
Hint
Which column proves a value instead of asking you to trust it?
What rule stops the model from computing a date you did not ask it to compute?
Why this is the answer
The assembled prompt names the columns, fixes one row per contract and adds the step that makes extraction verifiable: a Source quote column, so checking a value means reading one sentence. 'Copy, never compute' blocks the silent normalization that turns an extracted date into an invented one, and 'not stated' keeps gaps visible instead of filled.