Skip to content
ProtocolProtocol
← Your pathLevel 1 · Beginner

Finding the real header row, and the units row underneath it

Lesson 2 of 6 · Spreadsheets and instrument exports

1 · Learn the move · Constraints + negative instructions

Instrument exports bury the header: three rows of instrument name, method, and operator, then the column names, then - the trap inside the trap - a units row directly underneath, which a naive parse reads as the first data row. One row of text in a numeric column can silently change the column's type, and every statistic downstream inherits the damage. This is a job for negative instructions, because the failure mode is the model's default helpfulness: do not assume row 1 is the header; do not treat any row as data until the header and units rows are identified and shown to me; do not coerce a mixed column to make it parse; never invent a unit that is not written in the file. Constraints like these read as paranoia until the first time the mean of a concentration column comes back tenfold off because "ug/mL" was averaged in as zero.

Parse this instrument export under constraints. Do NOT assume row 1 is the header. Do NOT treat any row as data until you have shown me: the true header row number, the full preamble above it, and whether a units row sits below the header. Do NOT coerce mixed-type columns silently - report them. NEVER assign a unit not literally present in the file; a column with no stated unit is [NEEDS: unit]. Output the structure first, wait for my confirmation before any numbers. Export: [paste]

2 · Your turn. You write the prompt

Your plate reader's csv export starts with six lines of method summary, then column names, then a row reading RFU, sec, RFU, RFU across the measurement columns. Last month a labmate averaged that row into her data and her blank-corrected values were all slightly, consistently wrong. Write the prompt whose constraints make that error impossible.

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.