← all repositories
huangkiki/dailypaper-skills

Turn Claude Code into a tireless paper-reading intern

It exists because manually hunting through arXiv and HuggingFace every morning is a chore best delegated to an agent.

dailypaper-skills
Not currently ranked — collecting fresh signals.
star history

What it does This is a collection of Claude Code skills that turns conversational prompts into a full paper-processing pipeline. It pulls new submissions from HuggingFace Daily, Trending, and arXiv, scores them against your research keywords, and bins them into must-read, worth-reading, and skip piles. For the keepers, it generates structured Obsidian notes covering methods, experiments, formulas, and limitations, then updates directory pages and a concept index. It also watches GitHub trending repositories and can ingest papers from Zotero collections or local PDFs.

The interesting bit The author treats Claude Code less like a chatbot and more like a shell for long-running agent workflows. The entire pipeline—fetching, reviewing, note generation, and index maintenance—is driven by natural language, with the agent handling the drudgery of downloading figures and repairing broken external image links. A lightweight FastAPI web viewer is included for browsing recommendations and notes in a browser, and it even carries a built-in Claude dialog to trigger new runs.

Key highlights

  • Keyword-based scoring with positive, negative, and domain-boost filters to surface relevant papers.
  • Structured Obsidian output with backlinks, concept libraries, and auto-generated table-of-contents pages.
  • Multi-source ingestion: arXiv URLs, local PDFs, Zotero searches, and Zotero category trees.
  • Optional local web viewer for browsing daily recommendations, paper notes, and GitHub trending lists.
  • Conservative defaults: it writes Markdown into your vault but leaves git commit and push disabled by default.

Caveats

  • The README explicitly warns this is a personal workflow shared as-is, not a guaranteed stable product.
  • AI-generated recommendations, summaries, and notes may contain factual errors or misreadings; the author advises verifying content before using it in formal writing.
  • The workflow is tightly coupled to Claude Code and local Python tooling, so it is strictly a local-machine setup for now.

Verdict Ideal for researchers who already live in Obsidian and want an agentic assistant to handle daily literature triage. Skip it if you need a polished SaaS product or do not use Claude Code.

Frequently asked

What is huangkiki/dailypaper-skills?
It exists because manually hunting through arXiv and HuggingFace every morning is a chore best delegated to an agent.
Is dailypaper-skills open source?
Yes — huangkiki/dailypaper-skills is open source, released under the Apache-2.0 license.
What language is dailypaper-skills written in?
huangkiki/dailypaper-skills is primarily written in Python.
How popular is dailypaper-skills?
huangkiki/dailypaper-skills has 1.2k stars on GitHub.
Where can I find dailypaper-skills?
huangkiki/dailypaper-skills is on GitHub at https://github.com/huangkiki/dailypaper-skills.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.