What problem does it solve? Data readiness workflows stop at judgment calls (grain, PII, rollups, publish safety), but approvals given in chat are lost and cannot advance gates or be audited. This Skill closes the loop between a raised question and a recorded, committed human decision. ## Core Features & Use Cases - Decision Packaging: Renders a raised judgment call from unresolved-questions.md into a reviewable approval request with evidence, options, impact, and the required authority class, without ever answering it. - Decision Transcription: Writes a named human's answer into unresolved-questions.md (Resolution + status flip) and readiness-status.yaml (approvals[] entry with owner name and authority class), flipping a stage to pass only when required evidence already exists. - Refusal & Conflict Handling: Declines self-approval, wrong authority classes, phantom question IDs, duplicate requests, and silently overwriting prior approvals. - Use Case: A grain-confidence stop blocks a table's Gold Ready stage; the console packages the question for the data owner, then records their named answer so the gate can advance with auditable evidence. ## Quick Start Ask the agent to package the open judgment call in mappings/<table>/unresolved-questions.md into an approval request, then record the named human's decision back into the readiness artifacts.
