| name | paper-workflow |
| description | Daily paper discovery workflow: search arXiv + Semantic Scholar, score, deduplicate, and generate a recommendation note. Also handles individual paper searches by topic. |
| allowed-tools | Bash |
Paper Discovery Workflow
Find, score, and recommend academic papers matching the user's research interests.
Step 1: Choose the search mode
| Mode | When to use |
|---|
| Daily recommendations | User wants today's paper digest |
| Topic search | User asks about a specific topic |
| Conference search | User asks about papers from a specific venue/year |
Step 2: Run the search
Daily recommendations
daily_recommend(language="zh", dual_track=true)
This runs the full pipeline: arXiv recent papers + conference impact papers + arXiv innovation papers. Returns a daily note and a list of top papers for analysis.
Topic search
search_papers(query="chain of thought reasoning", top_n=10)
Conference search
search_conf_papers(venues="NeurIPS,ICML", year=2026, top_n=10)
Step 3: Present results to the user
Show the paper list with:
- Title, authors, arxiv_id
- Recommendation score
- Matched domain and keywords
Step 4: Deep analysis (if requested)
For papers the user wants to analyze in depth, follow the paper-analysis skill.
Notes
- Language setting comes from config (
zh or en). Default is Chinese.
skip_existing=true avoids re-analyzing papers already in paper-notes/.- The dual-track mode gives 2 conference papers (impact-ranked) + 2 arXiv papers (innovation-scored).
