What problem does it solve?
This Skill helps you turn an initial LaTeX paper draft into a stronger, publication-ready version by iteratively simulating peer review and applying targeted revisions while strictly enforcing accept/revert halt rules.
Core Features & Use Cases
- Simulated peer review: Generates structured strengths, weaknesses, questions, and axis scores using a reviewer rubric designed for conservative feedback.
- Targeted LaTeX revisions: Applies revisions guided by a verbatim Content Refinement Agent prompt while integrating reviewer questions into the manuscript text.
- Execution-grounded iteration control: Snapshots each iteration, recompiles with LaTeX, re-scores, and either accepts or reverts based on deterministic scoring/plateau logic to prevent score gaming.
Use case example: You have a draft in workspace/drafts/paper.tex and you want to improve scientific depth, technical execution, and evidence presentation by running 3 refinement iterations with real compile+score feedback, then promoting the best accepted snapshot to workspace/final/paper.tex.
Quick Start
Refine the current draft by delegating Step 5 when the orchestrator asks for peer-review-based content refinement, so it compiles each iteration, scores it, and promotes the best accepted snapshot.
