What problem does it solve?
Academic manuscripts often suffer from overclaiming, template-like AI phrasing, broken paragraph logic, and inconsistent terminology, and careless polishing can silently alter data, citations, or scientific meaning. This Skill restructures and refines existing text through a four-layer editing process that protects every number, comparison direction, and citation intent.
Core Features & Use Cases
- Fidelity-Protected Editing: Builds a fidelity ledger of values, directions, samples, and citation intent before rewriting, then verifies each item after revision.
- Four-Layer Revision: Diagnoses argument gaps, paragraph logic, sentence structure, and surface formatting in priority order rather than starting with synonym swaps.
- Style Audit Script: Runs
python scripts/style_audit.py <file> to flag overclaims, stock phrases, boosters, misuse of "significant", and dense transitions for human review.
- Use Case: A researcher translating a Chinese results section into English for Nature Communications uses the annotated mode to receive polished prose, key edit explanations, and flagged scientific risks without any change to reported statistics.
Quick Start
Use qinyan-nature-polishing to polish this abstract for a Nature-style journal in annotated mode without changing any data or conclusions.
