What problem does it solve? Working in Blue Yonder Luminate Planning, it is easy to confuse a safe scenario edit with a committing action like publishing a plan or releasing planned orders to ERP, and mistakes at that stage are expensive or irreversible. This Skill gives an AI agent the operator judgment to classify every planning action as read, sandboxed write, committing, or destructive before it runs. ## Core Features & Use Cases - Action risk classification: Distinguishes reads, private-scenario edits, baseline publishes, and releases to ERP, with explicit gating rules for each level. - Object and lifecycle model: Explains SKU-Location grain, planned vs firm vs released orders, scenarios, baselines, snapshots, time fences, and deployment RecShips. - Recovery playbooks: Provides step-by-step recovery patterns for bad publishes, released orders, and overwritten baselines, including when recovery crosses into ERP. - Use Case: A planner asks the agent to add 5,000 units of promotional demand at a DC. The agent forks a private scenario, runs net-change, verifies data freshness, and gates the publish and release steps for human approval instead of editing the baseline directly. ## Quick Start Ask the agent to raise the consensus forecast for a SKU at a specific location in Blue Yonder and explain which steps require approval before publishing.
