stripe-refund-auditor

To audit Stripe refunds for anomalies, run the audit script with --since and --until date arguments to pull per-refund data into a JSONL file, then run the grader script to compute an overall anomaly score and generate a markdown report highlighting refund-rate spikes, missing reasons, and duplicates.

Stripe refund anomaly detection surfaces refund-rate spikes, missing refund reasons, and duplicate refunds. The grader computes a deterministic anomaly score per refund and outputs a markdown report listing the highest-scoring anomalies with details like id, amount, currency, status, and reason.

Yes, auditing Stripe refunds requires a STRIPE_API_KEY set in the environment or a .env.local file, along with Python 3.10 or higher. Preflight and integration checks verify API key availability and ensure safe operation in Stripe test mode before executing the audit.

Generating a markdown report from Stripe refund data involves running the audit script to produce a per-refund JSONL file, followed by the grader script which computes deterministic anomaly scores and renders a human-readable markdown report focusing on the highest-scoring refund anomalies.

Comparing Stripe refund activity across periods is supported by the audit script, which accepts --since and --until date parameters to define the refund window. Running audits on separate date ranges produces individual JSONL files and anomaly scores for period-over-period comparison.

Stripe refund audit preflight checks enforce test mode to ensure safe operation, preventing unintended live transaction modifications during anomaly detection. This safeguard verifies the STRIPE_API_KEY environment configuration before the audit and grader scripts pull refund data and compute anomaly scores.