What problem does it solve? Restaurant managers asking "which items stopped selling?" often get a flat list of slow items with no diagnosis. This Skill distinguishes items that recently stalled from items that never sold, checks whether cancellations or bundle sales explain the pattern, and routes each item to a concrete action: replace, promote, remove, or fix operations. ## Core Features & Use Cases - Threshold-based classification: Labels each item as structurally dead (prior and current period sales both ≤3), stalled (prior ≥10 and current below 50% of prior), slowing, or insufficient sample, and prints the thresholds in every output. - Basket and cancellation cross-checks: Uses basket analysis to see if an item still moves in bundles, and cancellation data to catch items that were ordered but not fulfilled, redirecting those to operations fixes instead of menu changes. - Price-cut simulation for stalled items only: Runs a read-only reference scenario for stalled items to gauge markdown tolerance, explicitly flagged as not a demand forecast. - Use Case: A store manager asks "which menu items should I drop this month?" The Skill pulls dead-item candidates, classifies each by direction, checks bundle partners, and outputs a per-item action line handing promotions to the seasonal-promo skill and removals to the menu-editing skill. ## Quick Start Ask the assistant to find which menu items have stopped selling recently and recommend whether to replace, promote, or remove each one.
