A
licenseA
qualityB
maintenanceLets Claude run a small online store's back office — searching customers, looking up and listing orders, generating sales reports and low-stock alerts, updating order status, issuing refunds, and creating invoices — with server-side guardrails like refund caps, idempotency keys, an order state machine, and audit logging on every write. Includes a deterministic eval harness that grades agent behaviour against the actual database state.
8
MIT