What problem does it solve? AI-generated projects often ship with generic design patterns, broken builds, or fabricated results, and there is no objective gatekeeper deciding whether a project is actually done. This Skill acts as a strict design and engineering reviewer that blocks completion until quality standards are met. ## Core Features & Use Cases - Weighted Rubric Scoring: Scores six dimensions including frontend design (35%), engineering craft (20%), and functionality (15%), requiring a weighted total of 8.5/10 and frontend score of 9/10 for approval. - Hard Gates: Automatically rejects projects with failing builds, three or more compositional slop tells (centered layouts, gradient heroes, default Inter font), neon colors, or fabricated placeholder results. - Structured Verdict Loop: Emits a VERDICT schema with P0/P1 punchlist items referencing exact files and lines, plus recommended skills and MCPs, so a builder agent can iterate until approved. - Use Case: An orchestrator agent finishes a landing page and submits it for review; the judge opens it in Playwright, pulls computed styles, runs a slop scan, and returns REVISION with concrete fixes until the design passes. ## Quick Start Judge the project in the current directory against the stated goal and emit a verdict with a punchlist.
