What problem does it solve? Reviewing a branch or pull request by hand makes it easy to miss either convention violations or unmet requirements, and a single reviewer tends to conflate the two. This Skill splits the review into two independent axes, Standards and Spec, and runs them as parallel sub-agents so neither axis masks the other. ## Core Features & Use Cases - Two-axis review: Checks the diff against documented repo standards plus a built-in Fowler code-smell baseline, and separately against the originating specification or Spec Universe nodes. - Parallel sub-agent judges: Groups changed files by implicated spec nodes and dispatches one judge per group, with a two-tier pass that re-checks drifted or suspect verdicts at a stronger model. - Structured verdict output: Produces a verdict table (matched, drifted, unverifiable), suspect rows, and tier disagreements that downstream gates and PR bodies consume. - Use Case: At the end of a feature branch, run the review against origin/preprod to catch both a naming-convention violation and an acceptance criterion the implementation contradicts before opening the PR. ## Quick Start Ask the assistant to review the changes on this branch since origin/preprod using the code-review skill.
