Rewrite ML/CV/NLP papers into reviewer-friendly drafts - one-message-per-paragraph clarity, reverse outlining, claim-evidence alignment, clean teaser/pipeline figures, and adversarial self-review before submission.
---
name: research-paper-writing
description: Improve academic paper writing for ML/CV/NLP-style papers with clear section structure, paragraph flow, and reviewer-facing presentation. Use when drafting or revising Abstract, Introduction, Related Work, Method, Experiments, or Conclusion; polishing figures/tables; checking claim-support alignment; or self-review before submission.
---
Turn a paper into a high-clarity, reviewer-friendly draft prioritizing first impression, logical flow, and evidence-backed claims.
## Workflow
1. Clarify the paper story before sentence-level edits and build a mini-outline per subsection.
2. Rewrite paragraph-by-paragraph: one message per paragraph, stated in the first sentence, with self-contained nouns and clear sentence-to-sentence relations.
3. Run reverse outlining after each section - map every topic sentence to the thesis and every evidence point to its topic sentence; cut paragraphs that don't map.
4. Treat figures/tables as core content: clean teaser and pipeline figures, readable minimal-ink tables.
5. Check every major Abstract/Introduction claim against experimental evidence; weaken or remove unsupported claims.
6. Finish with an adversarial five-dimension self-review (contribution, clarity, experimental strength, evaluation, method soundness) and a claim-evidence map; load only the needed `references/<section>.md`.
Full skill & source: https://github.com/Master-cai/Research-Paper-Writing-Skills/tree/9ee5eddc10068cc52590b3a68a827d3a387f5af9/research-paper-writing