Skill Whatevertr Planilha Tidy — Development skill for Claude Code

whatevertr-planilha-tidy

*[Versão em português](README.pt-BR.md)*

A **skill for Claude** (Anthropic) that delivers **data spreadsheets** in **tidy** format — navigable by **humans and by other AIs** (*machine-readable*). One row per record, one column per attribute, no merged cells, real Excel Table, consistent types, and a **Schema** sheet another AI can read without guessing.

Part of the **NUD — Constellation Method**, by Thainá Ramos.

**Three names, one idea:** *tidy data* (the structure), *machine-readable / AI-navigable* (the property), *clean database* (the use).

---

What it does

• Turns raw input (messy spreadsheet, pasted data, text) into a tidy .xlsx: one row = one record, one column = one attribute.
• No merged cells in data, single-row headers (prefixes instead of multi-level: Time - car (years)), consistent types, ISO-8601 dates.
• Ships every workbook with a Premissas sheet (prose: sources, limits, methods) and a Schema sheet (tabular: column · type · unit · domain · example) — the machine-readable contract.
• Optional companions (--export-companheiros): .schema.json (structure) + .csv (data) for AI/pipelines that don't open .xlsx.
• Meaning always lives in the data, never only in color (a "status" is a column, not a red cell).

Principle: *if you stripped all formatting and exported to CSV, would the information still be complete?* If yes, it's AI-navigable.

Excel and Google Sheets

The principles are tool-neutral. The builder outputs `.xlsx`, which **imports cleanly into Google Sheets** (structure, colors, freeze, header all survive). The only Excel-specific object is the *Excel Table* (becomes a plain range in Sheets). Native Google Sheets generation is a possible future extension.

Structure

whatevertr-planilha-tidy/
├── SKILL.md                 # skill instructions
├── references/
│   ├── builder_patterns.md      # the spec format + builder.py API
│   ├── navegabilidade_ia.md     # why e