# Independent protocol review prompt

Review the attached protocol against its stated decision, source documents, data dictionary, and applicable current guidance. Work from the artifacts, not the drafting model's assurance. State what cannot be checked with the available evidence.

Identify material problems in the goal/design match, estimand or descriptive quantity, time-zero alignment, denominator, data observability, operational definitions, bias control, assumptions, uncertainty, sensitivity analyses, and feasibility. For ITC, inspect connectivity, transitivity, effect modifiers, outcome timing, shared evidence and marginal/conditional contrasts. For economics, separate budget impact from value and inspect input provenance. For randomized studies, inspect allocation, concealment, consent, harms, and monitoring.

Check citation existence and claim match, current guidance status, and the crosswalk's actual completeness. Look for unsupported specificity, outcome-informed choices presented as prespecified, invented data availability, or missing human decisions. A completed heading is not proof of an adequate design.

Return prioritized findings with the exact section, evidence, consequence, and proposed repair or decision. Separate errors from unresolved information and optional improvements. Give a disposition of ready for named human review, revision needed, or not assessable with supplied evidence. Do not approve, execute, or publish the study and do not invent missing facts to make it pass.