Repeatability
High
The transformation logic is structurally identical each run: parse a known format, map known field aliases to target column names, and write to a fixed output schema. Once the 12 vendor mappings are defined, every subsequent run is deterministic and repeatable.
Ambiguity Tolerance
High
Success criteria are crisp: a valid CSV with exactly five columns (product_id, vendor_name, sku, unit_cost, stock_qty) populated from all 12 feeds. An agent can verify completeness, row counts, and schema conformance programmatically with no subjective judgment needed.
Data & Tool Availability
High
The agent needs access to the 12 vendor feed files and write access to an output directory — both are straightforward to provide. No live APIs, credentials, or external systems are required beyond file I/O and standard parsing libraries.
Error Cost
Medium
A mapping error (e.g., swapping cost and stock_qty) could corrupt inventory data downstream, but the output is a CSV that can be reviewed before ingestion, making errors detectable and reversible before any real damage occurs. The risk is real but easily mitigated with a pre-load validation step.
Human Judgment Required
Low
The field alias mappings (e.g., 'cogs' → 'unit_cost') are unambiguous and can be enumerated upfront. There are no taste, ethics, or relationship dimensions to this task — it is pure structural transformation.