paper-reader

name	paper-reader
description	Use when asked to read, analyze, summarize, or generate a report for an arXiv or Hugging Face paper using native AI agent capabilities.

Paper Reader Skill (paper-reader)

This skill provides step-by-step instructions for reading, analyzing, auditing, extracting figures, and generating structured paper reports for arXiv and Hugging Face papers directly using native agent capabilities.

---

Output Target & Directory Structure

All generated outputs must be stored under the target directory: output/<YYYY-MM-DD>/<paper_id>/

Target structure:

output/<YYYY-MM-DD>/<paper_id>/
├── report.md
├── report.html
└── assets/
    └── figure1.png

(Where <YYYY-MM-DD> is current date and <paper_id> is the canonical arXiv ID, e.g. 2606.01075)

---

Execution Pipeline (6 Phases)

Follow this 6-phase agent workflow when triggered with a paper request (arXiv ID, arXiv URL, or Hugging Face paper link):

Phase 1: Searcher & Fetcher

1. Input Resolution:
   • Resolve input string/URL to canonical arXiv ID (e.g., 2606.01075 from https://arxiv.org/abs/2606.01075, https://arxiv.org/html/2606.01075, or https://huggingface.co/papers/2606.01075).
2. Fetch Metadata & Content:
   • Primary: Fetch HTML paper content via https://arxiv.org/html/<paper_id>.
   • Secondary / Fallback: Fetch abstract page via https://arxiv.org/abs/<paper_id> and PDF via https://arxiv.org/pdf/<paper_id>.pdf.
   • Extract title, authors, submission/publication date, primary arXiv subject category (e.g. cs.CV, cs.CL, cs.LG).

Phase 2: Classifier

Classify paper into exactly one of four structural types:

• Method / Algorithm: Introduces a new model architecture, loss function, algorithm, or training strategy.
• Benchmark / Dataset: Introduces a new dataset, benchmark suite, evaluation protocol, or dataset curation strategy.
• System / Infrastructure: Introduces a software framework, execution engine, distributed training system, or hardware efficiency optimization.
• Survey / Review: Provides a comprehensive literature review, taxonomy, analysis, or historical overview of a field.

Phase 3: Writer & Finder

• Writer Agent Role:
  • Synthesize a concise, high-impact TL;DR (1-2 sentences).
  • Summarize background and core motivation.
  • Provide a detailed technical methodology breakdown.
  • Extract and present key mathematical formulations using LaTeX syntax ($$...$$ block or \(...\) inline).
  • Summarize main empirical results and benchmark performance metrics.
• Finder Agent Role:
  • Identify 3-5 prior foundational papers build upon or compared against.
  • Explain key conceptual or architectural differences relative to these prior works.

Phase 4: Criticizer (Independent Audit)

Perform an objective critical audit highlighting limitations, unverified assumptions, or practical constraints. Categorize every audit item with a clear severity level:

• <b style="color:#dc2626">HIGH</b>: Fundamental bottleneck, unverified core baseline, theoretical weakness, or severe evaluation limitation.
• <b style="color:#d97706">MEDIUM</b>: Practical implementation constraint (e.g. extreme hardware/compute overhead, sensitive hyperparameter tuning, limited dataset diversity).
• <b style="color:#6b7280">LOW</b>: Minor scope boundary, missing non-critical ablation, or formatting/clarity artifact.

Phase 5: Figure Extraction

1. Locate the main overview diagram or Figure 1 from the arXiv HTML content (typically inside <figure> tags or image elements like x1.png, fig1.png).
2. Download or copy the image asset to output/<YYYY-MM-DD>/<paper_id>/assets/figure1.png.
3. Extract and save the corresponding figure caption text.
4. Fallback: If HTML image extraction fails, create a placeholder overview or extract figure caption reference.

Phase 6: Assembler

Generate both report.md and self-contained report.html in output/<YYYY-MM-DD>/<paper_id>/.

---

Output Templates & Schemas

1. report.md Template

# [Paper Title]

**Authors:** [Authors List]  
**ArXiv ID:** [arXiv:[paper_id]](https://arxiv.org/abs/[paper_id]) | **Date:** [YYYY-MM-DD] | **Category:** [Primary Category]  
**Paper Type:** [Method / Benchmark / System / Survey]  

> **TL;DR:** [One-paragraph concise distillation]

---

## Overview Figure

![Figure 1](assets/figure1.png)  
*Figure 1: [Figure Caption]*

---

## 1. Background & Motivation
[Context, problem formulation, and limitations of existing methods]

## 2. Methodology & Key Formulations
[Detailed technical breakdown]

$$
[Mathematical Equation]
$$

## 3. Experimental Results & Benchmarks
[Key quantitative findings, comparison tables, and baseline evaluations]

## 4. Critical Analysis (Audit)
- **[HIGH]** [Limitation 1]
- **[MEDIUM]** [Limitation 2]
- **[LOW]** [Limitation 3]

## 5. Related Papers
- [[Prior Paper Title]](https://arxiv.org/abs/XXXX.XXXXX) - [Brief relation and key distinction]
- [[Prior Paper Title 2]](https://arxiv.org/abs/XXXX.XXXXX) - [Brief relation and key distinction]

2. report.html Template (Self-Contained)

<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="UTF-8">
  <meta name="viewport" content="width=device-width, initial-scale=1.0">
  <title>Paper Report: [Paper Title]</title>
  <!-- MathJax JS -->
  <script src="https://cdn.jsdelivr.net/npm/mathjax@3/es5/tex-mml-chtml.js" async></script>
  <style>
    :root {
      --font-family: -apple-system, BlinkMacSystemFont, "Segoe UI", Roboto, Helvetica, Arial, sans-serif;
      --bg-color: #0f172a;
      --card-bg: #1e293b;
      --text-color: #f8fafc;
      --text-muted: #94a3b8;
      --accent-color: #38bdf8;
      --border-color: #334155;
    }
    body {
      font-family: var(--font-family);
      background-color: var(--bg-color);
      color: var(--text-color);
      line-height: 1.6;
      max-width: 900px;
      margin: 0 auto;
      padding: 2rem 1rem;
    }
    header {
      border-bottom: 1px solid var(--border-color);
      padding-bottom: 1.5rem;
      margin-bottom: 2rem;
    }
    h1 { color: #ffffff; font-size: 2rem; margin-bottom: 0.5rem; }
    .meta { color: var(--text-muted); font-size: 0.95rem; margin-bottom: 1rem; }
    .meta a { color: var(--accent-color); text-decoration: none; }
    .tldr-box {
      background: rgba(56, 189, 248, 0.1);
      border-left: 4px solid var(--accent-color);
      padding: 1rem 1.25rem;
      border-radius: 0.375rem;
      margin-bottom: 2rem;
    }
    section {
      background: var(--card-bg);
      border: 1px solid var(--border-color);
      border-radius: 0.5rem;
      padding: 1.5rem;
      margin-bottom: 1.5rem;
    }
    h2 { color: var(--accent-color); border-bottom: 1px solid var(--border-color); padding-bottom: 0.5rem; }
    .figure-container { text-align: center; margin: 1.5rem 0; }
    .figure-container img { max-width: 100%; height: auto; border-radius: 0.375rem; border: 1px solid var(--border-color); }
    .caption { font-size: 0.875rem; color: var(--text-muted); margin-top: 0.5rem; }
    .badge {
      display: inline-block;
      padding: 0.15rem 0.5rem;
      border-radius: 0.25rem;
      font-weight: bold;
      font-size: 0.8rem;
    }
    .badge-high { background: #7f1d1d; color: #fca5a5; }
    .badge-medium { background: #78350f; color: #fde68a; }
    .badge-low { background: #374151; color: #d1d5db; }
    code, pre { background: #0f172a; padding: 0.2rem 0.4rem; border-radius: 0.25rem; }
    ul { padding-left: 1.25rem; }
    li { margin-bottom: 0.5rem; }
  </style>
</head>
<body>

  <header>
    <h1>[Paper Title]</h1>
    <div class="meta">
      <strong>Authors:</strong> [Authors List]<br>
      <strong>ArXiv ID:</strong> <a href="https://arxiv.org/abs/[paper_id]">arXiv:[paper_id]</a> | 
      <strong>Date:</strong> [YYYY-MM-DD] | 
      <strong>Category:</strong> [Primary Category] | 
      <strong>Type:</strong> [Method / Benchmark / System / Survey]
    </div>
    <div class="tldr-box">
      <strong>TL;DR:</strong> [One-paragraph concise distillation]
    </div>
  </header>

  <main>
    <section>
      <h2>Overview Figure</h2>
      <div class="figure-container">
        <img src="assets/figure1.png" alt="Figure 1 Overview">
        <div class="caption">Figure 1: [Figure Caption]</div>
      </div>
    </section>

    <section>
      <h2>1. Background & Motivation</h2>
      <p>[Context and background problem statement]</p>
    </section>

    <section>
      <h2>2. Methodology & Key Formulations</h2>
      <p>[Technical explanation]</p>
      <div>
        $$[Mathematical Equation]$$
      </div>
    </section>

    <section>
      <h2>3. Experimental Results & Benchmarks</h2>
      <p>[Quantitative evaluation & metrics]</p>
    </section>

    <section>
      <h2>4. Critical Analysis (Audit)</h2>
      <ul>
        <li><span class="badge badge-high">HIGH</span> [Limitation 1]</li>
        <li><span class="badge badge-medium">MEDIUM</span> [Limitation 2]</li>
        <li><span class="badge badge-low">LOW</span> [Limitation 3]</li>
      </ul>
    </section>

    <section>
      <h2>5. Related Papers</h2>
      <ul>
        <li><a href="https://arxiv.org/abs/XXXX.XXXXX">[Prior Paper 1]</a> - [Brief relationship]</li>
        <li><a href="https://arxiv.org/abs/XXXX.XXXXX">[Prior Paper 2]</a> - [Brief relationship]</li>
      </ul>
    </section>
  </main>

</body>
</html>