Cleaning messy e-learning CSV data is exactly the kind of grunt work AI handles well.

Every transformation rule is explicitly stated and structurally identical across all 3,800 rows. This task could be run on a new export tomorrow with zero changes to the logic.

Success criteria are well-defined for each column: lowercase/trim emails, null future dates, numeric score conversion with flag, dedup by max score, and a clean_flag column. The only soft edge is 'obvious typos' in emails, which can be handled by regex heuristics and a flag rather than a hard decision.

The agent needs only the CSV file and a Python or pandas environment, both of which are standard and readily available. No external APIs, credentials, or live data sources are required.

The original CSV is preserved, the output is a new file, and the clean_flag column makes every transformation auditable. A human can spot-check and rerun if anything looks wrong with minimal consequence.

Pass/Fail-to-numeric conversion, deduplication logic, and date validation are all rule-based. Email typo detection is the one area requiring judgment, but flagging ambiguous cases for human review is a standard and acceptable fallback.