Prevent Duplicate Business Actions

The problem

Duplicate payments, orders, or system actions often result from race conditions, manual retries, or failed migrations. Developers and ops teams face the risk of issuing double refunds or stopping valid shipments when trying to clean up messy, multi-system data without a source of truth.

What it does

• Reconstructs action chains across checkout, payment, fulfillment, and support systems to identify the root cause of duplication.
• Maps business intentions against system events to distinguish between true duplicates and legitimate multi-stage actions like split shipments or repairs.
• Generates executable containment plans to pause retries, freeze bulk changes, and preserve logs during investigation.
• Designs durable prevention controls including idempotency keys, processed-event registers, and state-transition logic.
• Produces reconciliation registers that track financial, physical, and data outcomes for every affected record.

Why this beats prompting it yourself

General LLMs often suggest risky bulk deletions or reversals based on surface-level matches. This skill enforces strict reconciliation rules, requiring evidence of physical and financial outcomes before authorizing any irreversible action, preventing "double-correction" errors.

Use cases

• Recovering from a checkout outage where customers and support agents both re-ran failed orders.
• Reconciling duplicate purchase orders and receipts after a messy ERP data migration.
• Investigating webhook replays that triggered multiple notifications or account changes.
• Audit-ready cleanup of duplicate supplier invoices referencing different internal identifiers.

Known limitations

Requires logs or transaction histories provided via the Read tool. Does not execute API calls to reverse payments or delete database records directly.