What problem does it solve? Writing a peer review that is both thorough and disciplined is hard: reviewers either miss real errors or pad reviews with stock comments and preferences. This Skill produces reviews where every error is caught and every suggestion earns its place, in a consistent senior reviewer voice. ## Core Features & Use Cases - Systematic error sweep: Recomputes numbers, checks internal consistency, and covers statistics, machine learning, spatial methods, causal claims, reproducibility, and ethics using references/error_sweep.md. - Two-gate filtering: Errors pass unconditionally; suggestions are admitted only if a competent editor would not accept the paper without them, with worked admit/reject examples in references/filter_examples.md. - Voice and venue adaptation: Standardized verdict vocabulary, location-led comments with diagnosis plus fix, plain-text output for submission forms, and a distinct Brazilian Portuguese register for venues like ANPET via references/invariants.md. - Use Case: You receive a transportation journal manuscript to referee. The Skill reads the full paper, recomputes its tables, lists every error with locations and fixes, filters suggestions down to the load-bearing ones, and outputs a plain-text review with a standardized recommendation line. ## Quick Start Review this manuscript PDF and write my referee report for the journal, including a recommendation line and numbered comments.