Nature Paper X-Ray
Reconstructs how a paper's authors got their results, checking the evidence behind each claim
- Repository
- Yuan1z0825/nature-skills
- Stars
- 51,827
- Licence
- Apache-2.0
- Files
- 7 files
What it does
Once installed, when you give the agent one research paper, it reads the whole thing closely and writes a structured long-read explaining the path the authors took, which design choices actually mattered, and the worked math with concrete numbers. It also checks how far the evidence really reaches, flagging what the paper does not report instead of repeating the abstract. It is explicitly not a summary, translation, or peer review.
Great for
- Researchers who need to deeply understand a single paper before building on it or citing it
- Students preparing to present or explain a dense paper
- Reviewers or lab members who want to see the reasoning behind formulas and figures
Skip it if
- Quick summaries or translations: the skill explicitly declines these and produces a long Markdown analysis, which is slower and heavier than needed
- Judging whether a paper is publishable or novel: the skill sets no verdict and points to other skills for that
- Anyone expecting literature search or citation verification: those tasks belong to other skills named in the file
Try asking
- “精读这篇论文，讲透作者为什么这么设计 (paste or attach the PDF)”
- “这个公式怎么推出来的？请用具体数字举例说明”
- “Read this paper closely and check whether the baselines were tuned as hard as the proposed method”
- “Give me a paper X-ray of this attached study before I cite it in my review”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 7 original files, the Apache-2.0 licence and an attribution note.
nature-paper-xray.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip nature-paper-xray.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip nature-paper-xray.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Supply the full paper including appendix and supplementary material, since the skill relies on them for ablations and hyperparameter details.
- Ask in the language you want the output; Chinese requests produce Chinese prose with English technical terms kept.
- If the paper is long, ask the agent to state where its reading stopped, since the skill says it must not imply fuller coverage than achieved.
What's inside
- agents/openai.yaml
- manifest.yaml
- README_EN.md
- README.md
- references/deep-reading-protocol.md
- references/evidence-boundary.md
- SKILL.md
Redistributed unmodified from Yuan1z0825/nature-skills under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.