What problem does it solve? Academic authors need rigorous, referee-quality feedback on manuscripts before journal submission, but real peer review is slow and unpredictable. This Skill produces comprehensive manuscript reviews across six dimensions (argument, identification, econometrics, literature, writing, presentation) and can simulate an entire editorial pipeline calibrated to a target journal. ## Core Features & Use Cases - Three review modes: single-pass report for early drafts, --adversarial critic-fixer loop (max 5 rounds) that proposes and applies edits until approval, and --peer <journal> simulated peer review with an editor plus two dispositioned referees. - Journal calibration and variance analysis: referee dispositions are sampled from a 6-way taxonomy (STRUCTURAL, CREDIBILITY, MEASUREMENT, POLICY, THEORY, SKEPTIC), with --variance N producing a decision distribution, --stress running a hostile-editor gauntlet, and --r2/--r3 handling revise-and-resubmit rounds. - Cross-artifact verification: automatically invokes code review and reproducibility audits on referenced analysis scripts, and verifies editor novelty-probe claims before they enter the desk-review narrative. - Use Case: Before submitting to the QJE, run a simulated peer review with --peer QJE --variance 3 to get a decision distribution and concern-frequency table showing which criticisms are robust versus disposition-dependent. ## Quick Start Ask the AI to review your manuscript file, for example: review my paper at drafts/paper.tex in peer-review mode calibrated to AER with variance set to 3 referees.
