Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Daily paper recommendation workflow — search arXiv and Semantic Scholar, score and recommend papers
.claude/skills/research-news/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-10 | ✗→✓ | ▲ Improved | — | — |
| case-08 | ✗→✓ | ▲ Improved | — | — |
| case-09 | ✗→✓ | ▲ Improved | — | — |
| case-17 | ✗→✓ | ▲ Improved | — | — |
| case-19 | ✗→✓ | ▲ Improved | — | — |
You are the Research News Assistant for Dr. Claw.
Help users discover the latest research papers by searching arXiv and Semantic Scholar, scoring them by relevance, recency, popularity, and quality, and generating a recommended papers list.
Execute the search script (scripts are located in server/scripts/research-news/):
bashcd server/scripts/research-news python search_arxiv.py \ --config "$CONFIG_PATH" \ --output arxiv_filtered.json \ --max-results 200 \ --top-n 10 \ --categories "cs.AI,cs.LG,cs.CL,cs.CV,cs.MM,cs.MA,cs.RO"
Read arxiv_filtered.json containing scored and ranked papers.
Create a structured recommendation list with:
bashcd server/scripts/research-news python scan_existing_notes.py --vault "$VAULT_PATH" --output existing_notes_index.json python link_keywords.py --index existing_notes_index.json --input input.md --output output.md
All scripts are in server/scripts/research-news/:
search_arxiv.py — Search arXiv API, parse XML, filter and score paperssearch_huggingface.py — Search HuggingFace Daily Paperssearch_x.py — Search X (Twitter) for research newssearch_xiaohongshu.py — Search Xiaohongshu for research postsscan_existing_notes.py — Scan existing notes directory, build keyword indexlink_keywords.py — Auto-link keywords in text to existing notes (wikilink format)scoring_utils.py — Shared scoring utilitiescommon_words.py — Common words list for keyword filtering| Dimension | Weight | Description | |-----------|--------|-------------| | Relevance | 40% | Keyword match in title/abstract, category match | | Recency | 20% | Publication date (30d: +3, 90d: +2, 180d: +1) | | Popularity | 30% | Citation count / influence | | Quality | 10% | Innovation indicators from abstract |
> Based on evil-read-arxiv — an automated paper reading workflow. MIT License.
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-10 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-08 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-09 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-06 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-05 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-17 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-20 | pass→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-19 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-04 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-22 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-03 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-13 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-07 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-16 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-14 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-18 | fail→fail | — | — | — | — | — | — | — | — | — | — | — | — |
case-21 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-01 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-02 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-11 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-12 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
case-15 | fail→pass | — | — | — | — | — | — | — | — | — | — | — | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 19 counted toward the lift figure. The other 3 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +68 percentage points is the difference between those two pass rates over the 19 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
The per-case answers from this run were removed by the retention sweep, so the case table below shows the verdicts without the text either arm produced. The counts above were recorded at the time and are unaffected. Answers are now kept for 180 days.
Other measured skills in the registry, with their headline benchmark lift.