Data prompts

Act as a strict parser.

Source text:
[paste here]

Expected schema (JSON Schema or Zod):
[paste here]

Rules:
- Do NOT fabricate values.
- If a field is missing, return null and add a warning.
- Return ONLY valid JSON (no backticks).

Output:
{
  "data": <object matching schema>,
  "warnings": [<strings>]
}

Dataset (table or JSON):
[paste here]

Goal: normalize for analysis.
Rules:
- dates to ISO-8601 (YYYY-MM-DD) when possible; otherwise null + warning
- units: convert to [standard unit] (state assumptions)
- strings: trim + consistent casing
- deduplication: explicit criteria
Return:
1) normalized data
2) list of assumptions
3) ambiguity warnings

Input CSV:
[paste here]

I want to map it to this target schema:
[describe target columns + types + rules]

Do:
- define a mapping plan (col A -> col B)
- list cleaning rules per field
- detect missing columns and ask only what's necessary
- return the transformed CSV (with new headers)
If ambiguous: null + warning.

Act as a data analyst.

Requirement:
[what I want to measure]

Tables schema:
[paste tables + columns]

Deliver:
1) SQL (with comments)
2) assumptions (if info is missing)
3) quick tests: 3 queries to validate it’s not broken
Rule: if ambiguous, ask up to 3 clarifying questions before writing the final SQL.

JSON:
[paste here]

Business rules:
[paste here]

Task:
- check each rule and mark PASS/FAIL
- for each FAIL: exact path (e.g., data.items[2].price) + why
- propose minimal fix (patch)
Output in Markdown with a results table + JSON patch (RFC 6902 if applicable).

Text:
[paste here]

Extract entities into a table with columns:
- entity_type (person/company/product/date/amount/location/other)
- value
- confidence (0-1)
- evidence (exact quote)

Rules:
- evidence must be a literal quote from the text
- if unsure, confidence < 0.6
Return CSV with headers.