What problem does it solve? Operating Ivalua's source-to-pay suite is risky because its writes commit money and reach external suppliers, and every deployment is client-configured so behavior cannot be inferred from object types. This Skill provides the judgment to classify each Ivalua action (read / write / destructive) so consequential operations pause for human approval. ## Core Features & Use Cases - Action classification matrix: Maps Ivalua operation families (publishing RFx events, awarding scenarios, issuing POs, releasing invoices, changing supplier status) to read, reversible-write, committing-write, and destructive tiers with gating rules. - Configuration-as-control-surface rules: Treats no-code workflow, tolerance, and validation-rule changes as fleet-wide destructive changes, and defaults unclassifiable custom objects up to destructive tier. - Edge states and recovery: Covers sealed-bid envelopes, split awards, partial receipts, non-PO invoice auto-matching, ERP export failures, and prohibited circumvention patterns like threshold splitting. - Use Case: When asked to release an invoice with an out-of-tolerance match exception in Ivalua, the Skill flags it as destructive, requires a named approver, and suggests fixing the source receipt or price instead of overriding. ## Quick Start Ask the assistant to classify and safely process an Ivalua task, such as reviewing a match exception on a PO-based invoice before releasing it for payment.
