Nature Paper Xray
Read one supplied paper closely enough to reconstruct how its authors actually got...
Install
npx skills add Yuan1z0825/nature-skills --skill nature-paper-xray -g -a claude-code -y- Repository
- Yuan1z0825/nature-skills51.7k
- Category
- Design
- Updated
- 4d ago
- License
- Apache-2.0
- Context cost
- Lightabout 1,324 tokens
- Runs scripts
- No, instructions only
What it does
Read one supplied paper closely enough to reconstruct how its authors actually got there, rather than restating what the abstract claims. Use for 精读论文, 讲透这篇论文, 把这篇读懂, 作者为什么这么设计, 这个公式怎么推出来的, paper x-ray, and deep single-paper comprehension before building on, reviewing, presenting, or citing that paper. Recovers the real starting point and the bet the authors placed, separates load-bearing design from decoration, grounds each key formula in a worked micro-example with concrete numbers, and checks how far the evidence actually reaches, including where the authors are confident and where the paper does not report something. Delivers a structured Markdown long-read by default. Not a summary, not a translation, not a peer review.
Nature Paper X-Ray
Read a paper the way a researcher explains it at a whiteboard to a colleague: what the authors saw, why they believed it, where they hesitated, what they gave up, and which numbers would change the conclusion.
The delivered artifact is a Markdown long-read. A section-by-section summary of the paper is a failure mode, not a fallback.
Default stance
- The method section is a teaching order, not the order the work happened in. The introduction's story is arranged after the fact. The contribution list is written for reviewers.
- The judgments that decided the paper, what the authors noticed, why they trusted it, what they abandoned, are mostly not written down. Recovering them is the job.
- Every figure was chosen. Ask what the authors want a reader to believe from this one, then look for details inside it that do not support that reading.
- A figure contains its own counter-evidence more often than the text does: a curve that crosses later, an error bar wider than the gap between compared methods, a log axis hiding a constant-factor difference, one seed standing in for a variance claim.
- Name a mechanism or say the paper does not name one. Do not supply a motive the authors never stated.
