DailyPaper Skills
Interest-based paper discovery, reading priorities, and linked research notes from Hugging Face and arXiv.
Last reviewed by the paperbanana team on Oct 20, 2026
Install
git clone https://github.com/huangkiki/dailypaper-skills.git
cd dailypaper-skills
python3 install.py --agent codexWhat is DailyPaper Skills?
DailyPaper Skills connects recent-paper discovery with a personal reading workflow. Seven related skills cover Hugging Face and arXiv discovery, research-interest selection, candid reviews, priority-paper notes, concept links, indexes, and an optional GitHub-trending workflow. Its main daily-papers entry point runs discovery, review, and note generation as successive stages.
The current default ranking workflow uses the host agent to read candidate titles and abstracts and explain relevance to configured interests. A deterministic selection step validates the scores and selects qualifying papers, while history helps avoid repeats. Must-read papers continue into full reading notes; recommendations link to those notes and technical concepts.
Outputs are ordinary Markdown files; Obsidian and Zotero are optional. Installation does not create a scheduled task, and automatic Git commit and push are disabled by default. No additional model API key is needed for default agent ranking, but host-model usage still applies. Full PDF reading requires Poppler alongside Python, Git, and curl.
Core capabilities
Interest-based paper discovery
Fetch Hugging Face and arXiv candidates for a requested date range and use configurable interests, categories, and history to narrow the list.
Auditable relevance selection
Retain candidate scoring reasons and validate the complete score set before deterministic selection; relevance is distinct from scientific quality.
Reviews and priority notes
Explain what merits reading and send must-read papers through the full paper-reader workflow rather than relying on abstracts alone.
Connected Markdown knowledge base
Backfill links between recommendations, paper notes, concepts, and refreshed indexes, with Chinese and English output support.
What you can use it for
Catch up on a research area
Configure your interests and request today's papers or papers from the past week.
Read one paper in depth
Provide an arXiv link or a local PDF and request a full note, quick summary, or critical analysis.
Keep a reading archive navigable
Link new recommendations to existing notes and refresh paper and concept indexes after adding or moving material.
How to get started
- 1
Install the complete bundle
Run the clone and install.py commands above. Other documented targets include Claude, Cursor, Copilot, Gemini, OpenCode, and OpenClaw.
- 2
Set interests and folders
Run python3 skills/_shared/user_config.py --init from the cloned repository, then ask the agent to configure your interests and Markdown output directory.
- 3
Refresh the host session
Restart or refresh the agent so it can discover all seven skills and their shared resources.
- 4
Request and inspect a first run
Ask for today's papers, inspect the relevance reasons, and review the resulting recommendation file and priority-paper notes.
How it compares to similar skills
Choose DailyPaper Skills for a repeated discovery-to-reading workflow. It complements single-paper reading tools and broader research suites rather than replacing them.