What problem does it solve? AI agent work needs objective acceptance review, but reviewers often rewrite the work themselves, skip clauses, or invent evidence. This Skill enforces a strict rubric-based review where every clause of the done-when line gets an explicit pass or fail verdict. ## Core Features & Use Cases - Clause-by-clause grading: Splits the done-when line into separate clauses and answers pass or fail for each with a one-sentence justification. - Conjunction-based verdict: The work passes only if every clause passes; any fail or unknown fails the whole submission. - No-rewrite discipline: The reviewer never rewrites the work, never asks questions, and answers "Unknown" instead of inventing evidence when judgment is impossible. - Use Case: In a Todero agent orchestration pipeline, a judge agent reviews an engineer agent's handed-in code against the task's done-when line, listing each clause verdict before the state machine decides whether to accept or retry (up to two fail rounds). ## Quick Start Review the handed-in work against its done-when line, answering pass or fail on each clause and then one short paragraph explaining the verdict.
