ljg-skill-paper: The paper reader that refuses to stop at a summary
An open-source Claude Code skill that turns papers into judgments, maps, and durable notes.
**零术语规则**：承重概念必须场景化（3 级渐进），去掉技术名字仍能理解
- ljg-skill-xray-paper treats paper reading as critique and relevance mapping, not summary extraction.
- Its zero-jargon rule and cognitive collision cards make the paper's value visible against the reader's own notes.
- Org-mode output turns each pass through a paper into a durable knowledge object, not a disposable chat log.
- The repo is stricter than generic PDF chat on purpose, and that strictness is its product strategy.
Most paper tools try to compress. This one tries to clarify. It asks three different questions in sequence: what does the paper say, what is the gap, and what does it change in my world?
A reader built for judgment
The project comes from Li Jigang's Claude Code skill ecosystem, where plain language instructions do work that would otherwise live in code. The target user is someone who already keeps research in Emacs, Org-mode, and Denote, and wants the read to end as an artifact, not a chat transcript.
That is the first clue that this is not a normal summarizer. The repo behaves more like a reading protocol, with acquisition, deconstruction, and output chained into a repeatable path.
What makes it different
The sharpest move is the zero-jargon rule. Core ideas must be rewritten as scenarios, then stripped of technical names until the meaning still holds. On top of that, the skill adds cognitive collision cards: fork, tension, threshold, gap, or flip. The question is not just what the paper said. It is where it collides with what you already know, and whether the delta is real.
How the pipeline works
At the implementation level, the skill behaves like prompt-engineered software. It accepts several input types, normalizes them, fetches the source, then pushes the paper through a fixed sequence of reasoning operators. The output schema matters as much as the reasoning: a paper summary, a gap statement, a hypothesis check, an advisor verdict, and a machine-readable note structure.
#+filetags: :paper:review:
#+identifier: ljg-paper-xray
* 论文说了什么
** 关键结论
** 关键证据
* 对我意味着什么
** 可迁移方法
** 适用边界
* 博导审稿
** verdict: strong accept / weak accept / weak reject / strong rejectThat shape is the point. The repo treats formatting as cognition. If the note is easy to file, diff, and revisit, the reading becomes reusable instead of disposable.
Compared with generic paper tools
| Dimension | Generic PDF chat | ljg-skill-xray-paper |
|---|---|---|
| Primary job | Compress the paper into an answer | Force a judgment about meaning and gap |
| Output | Loose prose or section summaries | Structured Org-mode note and collision cards |
| Tone | Helpful and broad | Opinionated and strict |
| Best for | Quick digestion | Researchers who want a reusable reading artifact |
That makes it narrower than Elicit, Consensus, or a browser PDF assistant. But the narrowness is why it stands out. It is not trying to cover a literature review workflow. It is trying to turn the first read into a stronger second-order note.
值钱的不是 7 个 skill 本身，而是它把“理解内容 → 重写表达 → 视觉转译”做成了一套可拼装的工作流。
Why plain text still matters
The repo's ASCII diagrams and Org-mode output are not nostalgia. They are compatibility bets. Text survives terminal sessions, editor workflows, git diffs, and knowledge systems that outlive whatever front end is fashionable. In that sense, the primitive format is the product.
Seen that way, ljg-skill-xray-paper is a tiny manifesto. It says a good paper reader should not only summarize. It should clarify, challenge, and file the result where future-you can use it.