Repeatability
High
The task is structurally identical every month: same JSON schema, same transformation logic, same output format. This is a strong candidate for a reusable script that runs with zero modification.
Ambiguity Tolerance
High
Success criteria are crisp and enumerable: flat CSV, deduplication, avg_quiz_score, days_since_login, and an inactive flag at >30 days. There is one minor ambiguity around how to handle a student appearing in multiple courses (one row per enrollment vs. one aggregated row), but this is easily resolved by asking upfront or defaulting to one row per course enrollment.
Data & Tool Availability
High
The user exports the JSON files manually and can provide them directly to the agent. No live API access, authentication, or external system integration is required — just file I/O and standard data processing libraries.
Error Cost
Low
The source JSON files are untouched, so any mistake in the output CSV is fully reversible by re-running the script. The worst realistic outcome is a bad CSV that the user spots before acting on it.
Human Judgment Required
Low
Every step — parsing, flattening, deduplicating, calculating averages, computing date deltas, flagging inactives — is deterministic arithmetic and string manipulation. No taste, ethics, or relationship context is involved.