Prompt library
Document a dataset so others can trust it
Forcing a null-meaning and a gotcha entry per column is what turns a schema dump into documentation; those two columns hold everything analysts learn the hard way. The fitness statement sets the boundaries that prevent the dataset's numbers from wandering into claims they cannot support.
Last reviewed July 17, 2026
The prompt
Write documentation for the dataset below, for an analyst who has never seen it.
What I can tell you (schema, source, quirks I know of): {{dataset}}
Produce:
1. One-paragraph identity: what one row is, where rows come from, how often it updates, and the true freshness lag.
2. Column reference table: name, type, meaning, allowed values or range, null meaning (unknown? not applicable? zero?), and gotcha. The null-meaning column is mandatory per column; "no known gotcha" is an acceptable entry, silence is not.
3. Lineage: the systems and transformations upstream, and where a number in this table could diverge from the same number in the source system.
4. Known traps: the queries that look right and are wrong (timezone joins, soft-deleted rows, test accounts, pre-migration history), each with the correct pattern.
5. Fitness statement: what this dataset is authoritative for, what it is usable-with-caveats for, and what it must not be used for.
6. Open questions: everything I could not tell you, as a list for the data owner. Do not invent answers for these.
Write for skeptical reuse, not for compliance.Run in idaptOpens a new chat with the prompt prefilled. Nothing sends until you press send.
Fill in the variables
| Variable | What it is | Example |
|---|---|---|
| {{dataset}} | Schema plus what you know about it | [paste schema and your notes on where it comes from] |