DATA_CLEANUPData

Data Cleanup

Normalizes messy spreadsheets: dedupes, fixes types, and documents every transformation.

Membership required

// What this skill does

Data Cleanup profiles a raw CSV or spreadsheet, proposes a normalization plan, and applies it with a full audit trail. It handles inconsistent casing, mixed date formats, currency strings, near-duplicate records, and free-text categories. Every transformation is logged so the result can be reproduced or reversed.

// How to install

01Download the skill archive from this page.
02Unzip it and locate the SKILL.md file inside.
03Place the folder in your assistant's skills directory (or upload it to your project).
04Reference the skill by name in your prompt to invoke it.

// SKILL.md

Preview

# Data Cleanup

## Purpose
Normalize a raw tabular dataset with a reproducible audit trail.

## Inputs
- `dataset`: CSV or TSV contents
- `target_schema` (optional)

## Procedure
1. Profile every column: type, null rate, cardinality, sample values.
2. Propose transformations. Wait for confirmation on destructive steps.
3. Apply: type coercion, date normalization to ISO 8601, trim + case fold.
4. Detect near-duplicates with fuzzy matching. Report, never silently merge.

## Output
Cleaned dataset plus `## Transformation log` and `## Rejected rows`.