Repeatability
High
Data cleaning and normalization follows deterministic rules: strip currency symbols, convert units, parse text numerals, deduplicate by address. The structure is the same every run, making this highly automatable.
Ambiguity Tolerance
High
Success criteria are explicit: numeric prices, consistent units, deduplicated rows, flagged missing fields, CSV output. There is little room for subjective interpretation of what 'done' looks like.
Data & Tool Availability
High
The agent only needs the input spreadsheet file and standard data-processing libraries (pandas, regex, etc.). No external APIs or credentials are required.
Error Cost
Low
The original file is preserved and the output is a new CSV, so errors are fully reversible. A human spot-check of the output before use is trivial and catches any residual issues.
Human Judgment Required
Low
Parsing 'two bed' to 2 or converting sq meters to sq feet requires no taste or ethics — just pattern matching and arithmetic. Edge cases like ambiguous duplicates can be flagged for human review rather than silently resolved.