Data Cleaner

Every data export you get is broken in a different way.

Data Cleaner is an OpenClaw skill that takes a raw CSV, profiles it in seconds (types, null rates, suspected column roles), confirms the cleaning rules with you, normalizes dates to ISO 8601, phone numbers to E.164, country names to ISO 3166-1, deduplicates with a configurable strategy, and writes a clean file alongside a change log so you can audit every change made to the original.

No Python setup beyond the standard library. Runs on any CSV from Mercury, Chase, Amex, Wise, HubSpot, Salesforce — or whatever the ops team just emailed you.

What it does

• Profiles any CSV in seconds: per-column type, null %, distinct count, top values, and suspected role
• Normalizes dates across 8+ format variants to ISO 8601 (or your preferred format)
• Normalizes phone numbers from any local format to E.164 with configurable default country code
• Maps 80+ country name variants to ISO 3166-1 alpha-2
• Deduplicates rows by configurable key columns using first, last, or merge strategies
• Writes a change log — every original value and its normalized replacement, row by row

What's in the package

All files listed below are bundled as fenced code blocks inside SKILL.md — there's no separate folder to install. Copy each code block into its own file using the path shown above it.

• SKILL.md — the four-step workflow (profile, confirm, normalize, deliver) your agent follows.
• scripts/profile_csv.py — profiles a CSV with per-column type, null %, distinct count, top values, and suspected role detection.
• scripts/normalize_csv.py — applies normalization rules and writes a clean CSV plus a markdown change log.
• scripts/examples/ — sample messy CSV, profile output, clean CSV, and change log so you can verify it works on install.

How it ships

Drop SKILL.md into ~/.openclaw/skills/data-cleaner-csv/. Restart OpenClaw. Run /data-cleaner-csv or say "Clean this CSV" — your agent picks it up.

Questions: hackedcubeapps@gmail.com (response within 48 hours).