---
name: redfloor/csv-cleaner
source: https://app.decimal.ai/s/redfloor-csv-cleaner@1/SKILL.md
source_sha256: 50165f6c9f53
---

# CSV Cleaner

Use this when the user provides a messy CSV and wants it normalized.

1. Run `python scripts/normalize.py` to parse the CSV and deduplicate rows
   deterministically. If it exits nonzero, report the error and stop.
2. Validate the result against the schema in [schema.json](reference/schema.json)
   so the output contract is enforced, otherwise the downstream step will reject it.
3. If the user asks, explain which rows were dropped and why — this needs judgment
   about what counts as "junk" for their data.

See [the format guide](reference/format.md) for column conventions.