paper-onion

Paper Onion Peeling Paper Onion

Core Concept

Reading a paper is not information dumping, but a process ofprogressive internalization. This skill simulates how an experienced researcher truly understands a paper:

1. First spend 30 seconds scanning it to build a first impression
2. Then spend 5 minutes clarifying the skeleton of the storyline
3. Then spend 20 minutes going deep into the core mechanism
4. Finally spend 10 minutes internalizing it into your own understanding and connecting it with existing knowledge

Each layer is not simply information extraction, but answering acognitive question。

Input Processing

Accept the following input types:

• PDF file path (read from /mnt/user-data/uploads/)
• Paper URL (requires fetch)
• Directly pasted paper text

If it is a PDF, first extract the full text. If extraction is difficult, make the best effort to analyze based on the available parts (title, abstract, figure and table descriptions), and inform the user.

Four-Layer Onion Peeling Process

Perform the following four layers of analysis on the paper. Each layer is driven by acognitive question, rather than mechanically filling out a form.

First Layer · 30-Second Scan (👁)

Cognitive question:What is this paper saying?

Look only at the title, abstract, first paragraph of the introduction, and conclusion, and extract:

• one_sentence: One-sentence summary (no more than 40 characters, in Chinese, and it must be understandable even to non-experts)
• key_question: The core question this paper attempts to answer (expressed as a question)
• first_impression: 3 first-impression key points (each no more than 15 characters)

Style requirement: As natural as saying to a friend after a 30-second scan, "This paper is about..." Do not use academic jargon.

Second Layer · Skeleton Perspective (🦴)

Cognitive question:What is the logical skeleton of the paper?

Quickly read through the full text and extract the paper's "storyline":

• storyline: 4-step storyline, each step containing label and text
  • The label for step 1 uses "pain point": What is the key bottleneck of previous methods
  • The label for step 2 uses "inspiration": What is the author's breakthrough insight
  • The label for step 3 uses "solution": How it is specifically done (summarize the method in one sentence)
  • The label for step 4 uses "evidence": Key experimental data (must include specific numbers)
• key_figure: A description of the most important figure/table in the paper (specify which Figure it is and clearly explain what it shows)

Style requirement: The storyline should be like a narrative of "previously → discovered → did → proved", not a dry list.

Third Layer · Deep Dissection (🔬)

Cognitive question:How exactly does the core mechanism work?

Carefully analyze the methods section and extract:

• mechanisms: 3-5 core mechanisms, each containing:
  • what: Mechanism name
  • why: Why it is designed this way (not "what it is", but "why")
  • formula: Key formula or pseudocode (use concise mathematical notation; if there is no formula, describe the algorithm flow in one sentence)
• hidden_assumptions: 2-3 assumptions the paper does not state explicitly but implies (this is the most valuable part—what the authors don't say but you should know)

Style requirement: each why must answer "what would happen if this were not done?" Invisible assumptions should be sharp but fair.

Layer 4 · Internalized Connections (🧠)

Cognitive question:What does this mean for me?

This is the most important layer, and also the one most people skip:

• napkin_formula: Napkin formula—condense the paper's core insight in no more than 10 words (an extremely compressed expression like "parallel attention > sequential recurrence" or "diffusion = reverse denoising")
• core_insight: Core insight—2-3 sentences explaining what cognition this paper changed. It is not a repetition of the abstract, but "what did you understand after reading it that you did not understand before"
• connections: 3-5 knowledge connections, each containing:
  • to: Other papers/concepts/methods it connects to (write the names and years clearly)
  • link: What relationship this connection is (inheritance/contrast/complementarity/inspiration/generalization), explained clearly in one sentence
• what_changed: "How my cognition changed" after reading this paper—written in first person, like a diary

Style requirement: this layer best demonstrates depth of understanding. If the connections you can think of are only the references cited by the paper itself, your understanding is not deep enough—connect across domains.

Output: React Artifact note card

After completing the four-layer analysis, generate a React artifact (.jsx), presented as a hand-drawn style note card.

For visual style and implementation details, please read references/notecard-template.md。

Key design principles

1. One card condenses all the essence: It is not a simple concatenation of the four layers, but an information architecture with primary and secondary elements
2. Hand-drawn/paper texture: warm paper background color (#fdf8ef), handwritten font (Caveat), slightly tilted elements, dashed borders
3. Napkin formula centered and enlarged: This is the visual focus of the entire card
4. Storyline connected with arrows: pain point → inspiration → solution → evidence, clear at a glance
5. Knowledge connections use colored left borders: one color per connection, like branches of a mind map
6. Invisible assumptions use a warning color: places to remind the reader to pay attention to

Interaction method

The output of this skill isgenerate the complete note card in one go(not an interactive multi-step operation). But during generation, organize the thinking in the order of the four layers to ensure each layer is deep enough.

If the user asks follow-up questions about the details of a certain layer after viewing the note card, you can expand the discussion, but there is no need to regenerate the entire card.

Language

• All analysis content usesChineseoutput
• Paper titles, names of people, and terminology retain the original English
• Formulas use English/mathematical symbols
• Style: colloquial, warm, avoid academic clichés. Imagine you are explaining this paper to a smart friend in a café