Inventory / Stock Reconciliation

The skill recommends the smallest useful inventory and stock reconciliation setup for a business. Its default output is a short recommendation; CSV, SQL DDL, JSON Schema and a Notion property mapping are built only when requested. A single field list keeps those representations aligned.

The process treats counting, verifying and approving as three separate responsibilities. Counted By identifies the person who counted stock, Verified By identifies the verifier, and Approved By identifies the adjustment approver. One role is never inferred from another. Combining responsibilities requires an explicit statement about the business's actual process.

Use it to structure periodic stock counts and reconciliation responsibilities. Count frequency is a business decision based on volume, value and risk. Book quantity reflects recorded entries; an actual shelf count establishes what is present.

A correction booked without authorization can amount to a quiet write-off. Missing cartons might reflect theft, a mis-pick or two dispatches that never received a stock entry; the process leaves that investigation to a person. The manager approving an adjustment is not automatically its verifier, and the verifier is not automatically its counter.

Every adjustment needs a named approver before booking. The template does not replace human investigation of shrinkage: determining whether missing stock reflects theft, a picking error or unrecorded dispatches requires judgment rather than a formula.

Requested artifacts are empty structures. A CSV contains a header row without data rows; SQL DDL contains no inserts; the schema contains no example values; and the Notion output is a property mapping table. A clearly labelled illustrative row is added only on explicit request. Unknown information must remain unknown instead of turning into invented data inside an export.