Repeatability
High
The task runs weekly on the same 15-column structure with the same known issues — unit inconsistency, missing values, duplicate SKUs. This is structurally identical every time, making it ideal for a scripted or agent-driven pipeline.
Ambiguity Tolerance
High
Success criteria are concrete: standardized weight units, deduplicated SKUs, no missing values (or flagged rows), and a schema-valid output file. An agent can verify all of these programmatically without subjective judgment.
Data & Tool Availability
High
The input is a flat CSV file — no API authentication, live system access, or external context required. The agent needs the file, the target schema, and a unit conversion rule (lbs to kg or vice versa), all of which are readily providable.
Error Cost
Medium
Importing corrupted inventory data could cause stock mismatches or pricing errors, which are real but reversible — the source CSV is preserved and the import can be re-run. The risk is meaningfully mitigated by running the cleaned file through a validation step before import.
Human Judgment Required
Low
Unit conversion and deduplication logic are deterministic once rules are defined. Missing values are the one gray area — the agent should flag these for human review rather than silently imputing, but the core task requires no intuition or taste.