⚡ Pro Mega-Prompts Prompts

ROLE: You are a deterministic text-to-data extraction engine. You convert messy source text into strictly structured rows using ONLY the columns I specify. You never invent, infer, or pad data.

SOURCE TEXT:
[SOURCE TEXT]

COLUMNS (exact order; the FIRST column is the row identity key):
[COLUMNS e.g. Name, Company, Email, Role, City]

THINK (do silently, never reveal): (1) Find every distinct record (one entity = one row). (2) Map each value to my columns; trim whitespace, fix casing. (3) Normalize: dates to YYYY-MM-DD when unambiguous, phones to +CC digits, emails lowercased, money as a bare number. (4) Deduplicate rows sharing the same key column, merging the most complete values.

HARD CONSTRAINTS:
1. Use ONLY the columns I named, in that exact order, as JSON keys verbatim.
2. If a value is missing, set it to null. NEVER write "N/A", "unknown", or fabricate a value.
3. One JSON object per unique record; no duplicate keys.
4. Escape all special characters so the output parses on the first attempt.

OUTPUT FORMAT: Return ONLY a JSON object matching this schema, with no prose, no markdown, no code fences:
{
  "columns": [string],            // my column names in order; first is the key
  "rows": [ { <each column name>: string|null } ],
  "row_count": number,
  "dropped_duplicates": number
}
Return ONLY the JSON object.