Repeatability
High
This is a canonical software task with a well-understood structure: read CSV, deduplicate, handle errors, log, export. The pattern is nearly identical across instances, with only minor variation in column names or deduplication logic.
Ambiguity Tolerance
High
Success criteria are crisp: the script runs without errors, removes duplicates, logs events, and exports clean output. A human reviewer can verify correctness by running the script against test data.
Data & Tool Availability
High
The agent only needs a Python environment and optionally a sample CSV to test against. No external APIs, credentials, or live systems are required to produce the script.
Error Cost
Low
The deliverable is a code file that can be reviewed, tested, and corrected before touching any real data. Mistakes are easily caught and reversed with no downstream damage.
Human Judgment Required
Low
Deduplication logic, error handling patterns, and logging conventions are well-established in Python; no subjective taste or domain intuition is needed. A human should review the output but is not required to produce it.