uv-read-arxiv-paper

📁 uv-xiao/pkbllm 📅 1 day ago
3
总安装量
2
周安装量
#62291
全站排名
安装命令
npx skills add https://github.com/uv-xiao/pkbllm --skill uv-read-arxiv-paper

Agent 安装分布

opencode 2
gemini-cli 2
claude-code 2
github-copilot 2
codex 2
kimi-cli 2

Skill 文档

Read an arXiv paper (deep report)

Output locations (pkbllm convention)

  • Tracked output (write here): $HUMAN_MATERIAL_PATH/research/<paper_slug>/
    • report.md (main deliverable)
    • figures/ (optional images/diagrams)
  • Local downloads (gitignored, configurable): $HUMAN_MATERIAL_PATH/.references/
    • PDFs default: .references/pdfs/
    • Sources default: .references/arxiv/
    • Cloned repos default: .references/repos/

Config precedence (repo overrides user):

  1. $HUMAN_MATERIAL_PATH/.agents/config.toml
  2. ~/.agents/config.toml

Quick start

  1. Download paper assets:
# Run from this skill directory (the folder containing this SKILL.md):
python scripts/download_arxiv.py https://arxiv.org/abs/2601.07372
  1. Extract the TeX source:
python scripts/extract_arxiv_source.py 2601.07372
  1. Write the report:
  • Create $HUMAN_MATERIAL_PATH/research/<paper_slug>/
  • Start from assets/report_template.md
  • Fill it with a coherent story:
    • What problem is being solved and why now?
    • What is the key trick (the one idea that makes it work)?
    • What are the assumptions and failure modes?
    • What to implement first if we were to reproduce it?

Workflow (recommended)

  1. Normalize the paper id from the URL (keep the version if present; it’s fine).
  2. Download PDF + source into .references/ using the provided scripts.
  3. Unpack sources into .references/arxiv/<arxiv_id>/ and locate the LaTeX entrypoint (main.tex, etc.).
  4. Read in passes:
    • Pass 1: abstract + intro + figures (build the mental model)
    • Pass 2: method (write down the algorithm precisely)
    • Pass 3: experiments (what matters, what doesn’t)
    • Pass 4: limitations + related work (what breaks, what’s missing)
  5. Produce a report under $HUMAN_MATERIAL_PATH/research/<paper_slug>/report.md:
    • Use the template (copy it and fill it)
    • Prefer concrete examples, small equations, and “why this design” explanations
    • Include at least one visual explanation if it helps (optionally generate via uv-scientific-schematics)