A few short questions, plain English, no jargon. Your answers become Claude's standing instructions.
Where does this data come from? Examples: "CRM export (CSV)", "PostgreSQL events table", "Google Sheets survey results", "API JSON dump".
Describe every data quality problem you have already noticed. The more specific, the better. Examples: "dates in three formats", "~5% of rows have a null email", "country names are inconsistent (US, USA, United States)".
What happens after cleaning? This determines how strict the cleaning needs to be and which issues are blocking vs. acceptable.