What problem does it solve?
Writing and polishing SCI journal papers in LaTeX is slow and error-prone: numbers can be fabricated or misplaced, sections drift from their intended roles, abbreviations become inconsistent across files, and AI-generated text often reads like a generic template rather than the author's voice.
Core Features & Use Cases
- Dual run modes: autonomous mode directly edits the manuscript's .tex files, while collaborative mode only produces a review plan in plans/ for human approval before any change.
- Style-aware writing: built-in bensz-01 (biomedical, author-like) and general-01 (general SCI) styles, plus the ability to distill style signals from user-provided reference papers without copying sentences.
- Multi-layer review pipeline: parallel number verification, section-role checks with a dedicated Discussion audit, full-manuscript abbreviation consistency checks, and iterative logic-tree reviews before rendering PDF/Word output.
- Use Case: Point the skill at a LaTeX paper directory with figure/table notes, and it rewrites the Discussion to stop repeating Results, verifies every statistic against source materials, unifies abbreviations across all .tex files, and rebuilds the PDF.
Quick Start
Ask the AI to use the paper-write-sci skill to optimize your SCI paper, providing the LaTeX project directory and a figure/table notes file, optionally specifying collaborative mode or a writing style.