Best AI Skills for Scientific & Research in 2026: 15 Compared
Top 15 research and science skills for Claude Code. Nine are web search; only three (Biomni, AnnData, ClinVar) are actual scientific tooling — and they're the only 3/5-quality picks.
The Scientific & Research category is structurally distorted. The catalog has 317 skills here, but when you rank by usage, 9 of the top 15 are web-search engines and Weather at #1 isn't a scientific skill at all. The actually-scientific tooling — bioinformatics, genomic databases, scientific data analysis — sits at #9-#11 and represents the only entries in the category carrying a quality score (3/5, the only ones rated). If you came here looking for "what skill should I use for science work?", the answer is not the most-installed entry. We ranked the top 15 to make that visible.
Quick Pick
Biomni — Bioinformatics + research skill from K-Dense-AI's curated scientific-skills bundle. The first entry on this list that's actually built for science, with a 3/5 quality score earned through editorial review. Rank #9 by raw signal, but rank #1 if you're asking "which scientific skill should I install first." Comes packaged alongside #10 (AnnData) and #11 (ClinVar) — install the bundle.
What These Skills Actually Do
The Scientific & Research category clusters into three patterns, and the install signal is dominated by the first one: (1) Web search engines (#2–#7, #14, #15 — nine entries) — Multi-engine search, Baidu, Tavily, Brave, etc. These are useful for literature search and citation hunting but they're not science-specific; the same skills appear in any researcher's workflow. (2) Information aggregation (#8 YouTube, #12 News, #13 RSS, #1 Weather) — these pull external sources into Claude's context. (3) Actual scientific tooling (#9 Biomni, #10 AnnData, #11 ClinVar) — bioinformatics, single-cell genomics, and clinical-variant databases. Only three skills in the entire top 15. The pattern reveals what Claude Code is currently used for in "research" workflows: mostly searching the web with extra steps. Real science skills exist (and the ranking pushes them up via quality score) but the broader audience is still doing background research, not bench science.
How We Ranked
We sorted 15 candidate skills by a composite score:
- Popularity signal — the highest of GitHub stars, install count, or ClawHub download count. Log-scaled so a 100-star skill doesn't get buried under a 100,000-star one if the smaller one is meaningfully better.
- Quality score — when set, a 0–5 rubric that breaks ties within popularity tiers. Roughly 15% of catalog skills carry a quality score today; we surface it in the comparison table when available.
The formula is identical across the entire Best-Of 2026 series, so you can compare apples to apples between categories.
The Top 15
1. Weather
Skill · clawhub/weather · 124.0K signal · quality unrated Get current weather and forecasts (no API key required).
The take: A weather skill at the top of "Scientific & Research" tells you everything about how this category gets used. The skill itself is fine (no API key, decent forecast accuracy), but its #1 rank here is essentially a tag-misfit. If you do meteorological work, install it; otherwise skip and go to #9.
2. Multi Search Engine
Skill · clawhub/multi-search-engine · 102.0K signal · quality unrated 17-engine search aggregation (8 CN + 9 Global). Advanced operators, time filters, site search, privacy engines, WolframAlpha. No API keys.
The take: The most comprehensive search wrapper in the category — 17 engines, no API keys required. The CN/global mix is the differentiator over Prismfy (#3) and Tavily (#5, #14, #15). Install this or one of the others; running multiple search wrappers wastes context budget.
3. Prismfy Web Search
Skill · uroboros1205/prismfy-search · 77.5K signal · quality unrated 10-engine search via Prismfy: Google, Reddit, GitHub, arXiv, Hacker News, and more. Free tier included.
The take: The arXiv inclusion is the only thing that makes this "scientific." If you're doing academic literature search, Prismfy's arXiv integration beats the generic alternatives. For everything else, #2 has wider coverage.
4. Baidu Search
Skill · clawhub/baidu-search · 71.3K signal · quality unrated Search via Baidu AI Search Engine (BDSE). Live information, documentation, research topics.
The take: China-market essential. Baidu indexes Chinese-language web content far better than Google does, so for research with Chinese sources this beats the global engines. Skip for English-only work.
5. Tavily 搜索
Skill · jacky1n7/openclaw-tavily-search · 70.8K signal · quality unrated Web search via Tavily API — LLM-optimized; Brave alternative.
The take: First of three Tavily wrappers in the top 15 (#5, #14, #15). Tavily's pitch is "LLM-optimized" — snippets and metadata structured for AI consumption rather than human reading. Marginal advantage over Brave (#6) on returned content quality; significant disadvantage in that it requires a paid API key.
6. Brave Search
Skill · steipete/brave-search · 47.6K signal · quality unrated Web search + content extraction via Brave Search API. Lightweight, no browser required.
The take: Brave's API is the privacy-focused alternative to Google. The "no browser required" part matters more than it sounds — many search skills launch a headless Chromium, this one just hits an API. Faster, lower-resource. Paid API key required.
7. baidu web search
Skill · clawhub/baidu-web-search · 46.0K signal · quality unrated Baidu Qianfan web search API — real-time retrieval, news, fact verification.
The take: Alternative implementation of #4 using Baidu's Qianfan API rather than the consumer BDSE. Qianfan is Baidu's LLM-enterprise endpoint, so this is the more rate-stable choice for production workflows. Same caveat: most useful for Chinese-language sources.
8. YouTube Watcher
Skill · michaelgathara/youtube-watcher · 39.0K signal · quality unrated Fetch and read YouTube transcripts. Summarize, Q&A, information extraction.
The take: Critical research primitive — a huge amount of expert content lives in YouTube videos that never make it to text-indexed sources. Skill is simple (transcript fetch, parse, return), but the workflow it unlocks is meaningful: "read this conference talk for me" replaces 45 minutes of viewing with 30 seconds of querying.
9. Biomni
Skill · K-Dense-AI/claude-scientific-skills · 17.9K signal · quality 3/5 Bioinformatics + research skill from K-Dense-AI's scientific-skills bundle.
The take: The first actual science skill on the list, and one of only three carrying a 3/5 quality score in the entire category — earned through editorial review. Biomni focuses on biomedical workflows (sequence analysis, pathway exploration, literature linking). The quality score bumped it above several 30K+ signal skills via the ranking formula's tiebreaker, which is the whole point of having a quality rubric.
10. AnnData
Skill · K-Dense-AI/claude-scientific-skills · 17.9K signal · quality 3/5 Single-cell genomics data structure (AnnData/scanpy) skill from the same K-Dense-AI bundle.
The take: AnnData is the standard data structure for single-cell genomics work (RNA-seq, ATAC-seq). This skill teaches Claude the library's idioms — slicing, filtering, dimensionality reduction. Niche, but essential if you do bench bioinformatics; otherwise irrelevant.
11. ClinVar Database
Skill · K-Dense-AI/claude-scientific-skills · 17.9K signal · quality 3/5 Clinical-variant database access (NCBI ClinVar) from the K-Dense-AI scientific-skills bundle.
The take: ClinVar is the public database of clinical-significance variants — the gold-standard source for "is this mutation pathogenic." Skill enables direct query from Claude. Use case is narrow (clinical genetics) but in that use case, this is the skill. Quality-rated because the workflows it enables are too high-stakes for a generic search skill.
12. News Summary
Skill · joargp/news-summary · 33.7K signal · quality unrated RSS-based daily news briefings with optional voice summaries. Trusted international feeds.
The take: Curated RSS + summarization, with voice output as a nice-to-have. The curation is the value (someone vetted "trusted" feeds), since you skip the work of building your own feed list. Useful for daily-briefing workflows; less useful for one-off research.
13. Blogwatcher
Skill · steipete/blogwatcher · 31.6K signal · quality unrated Monitor blogs and RSS/Atom feeds for updates via the blogwatcher CLI.
The take: Lower-level alternative to #12 — gives you the RSS monitoring without the curation or summarization. Install if you already know which feeds you care about and want raw "new post" notifications routed through Claude. Pair with a writing skill for the summarization layer.
14. Tavily Search
Skill · matthew77/liang-tavily-search · 28.0K signal · quality unrated Tavily API wrapper — relevant results with content snippets, scores, metadata.
The take: Functionally near-identical to #5 (same Tavily backend). The redundancy in this list — three Tavily wrappers in the top 15 — is itself a data point about how undifferentiated the search-wrapper ecosystem is.
15. Tavily AI Search
Skill · bert-builder/tavily · 27.7K signal · quality unrated Third Tavily wrapper — AI-optimized search, comprehensive research, current events, domain-specific lookup, answer summaries.
The take: Pick one Tavily skill (#5, #14, or this one), not three. The marginal differences are which extra features each author bolted on; the underlying engine is identical. If you only do English-language research, this one's "AI answer summary" feature is the most useful extra.
Comparison Table
| # | Skill | Type | Stars / Installs | Quality | License |
|---|---|---|---|---|---|
| 1 | Weather | Skill | 124.0K | — | MIT-0 (Free to use, modify, and redistribute. No a |
| 2 | Multi Search Engine | Skill | 102.0K | — | MIT-0 (Free to use, modify, and redistribute. No a |
| 3 | Prismfy Web Search | Skill | 77.5K | — | — |
| 4 | Baidu Search | Skill | 71.3K | — | MIT-0 (Free to use, modify, and redistribute. No a |
| 5 | Tavily 搜索 | Skill | 70.8K | — | — |
| 6 | Brave Search | Skill | 47.6K | — | — |
| 7 | baidu web search | Skill | 46.0K | — | MIT-0 (Free to use, modify, and redistribute. No a |
| 8 | YouTube Watcher | Skill | 39.0K | — | — |
| 9 | Biomni | Skill | 17.9K | 3/5 | — |
| 10 | AnnData | Skill | 17.9K | 3/5 | — |
| 11 | ClinVar Database | Skill | 17.9K | 3/5 | — |
| 12 | News Summary | Skill | 33.7K | — | — |
| 13 | Blogwatcher | Skill | 31.6K | — | — |
| 14 | Tavily Search | Skill | 28.0K | — | — |
| 15 | Tavily AI Search | Skill | 27.7K | — | — |
FAQ
How is this list different from the category page on aiskill.market?
The category page is a directory: every skill in the category, sortable and filterable. This list is editorial — opinionated, time-stamped (2026-05-18), and ranked. Use the directory when you know what you want; use this when you don't.
Why does the #1 pick have fewer stars than #5?
Stars are one signal among several. The composite score above also includes install counts (which reflect actual usage on aiskill.market) and the optional quality score. A skill with a smaller star count can rank higher if its installs or quality score are strong enough to offset.
Are these all free?
The skills themselves are open source. API-backed search skills require keys: Tavily (#5, #14, #15) and Brave (#6) charge per query above a small free tier. Baidu Qianfan (#7) requires a Baidu Cloud account. The bioinformatics skills (#9, #10, #11) hit public databases (NCBI ClinVar is free; AnnData operates on your local data) — no API costs.
How do I install one?
Each linked skill page has install instructions. The fastest path is the one-line install via the aiskill.market CLI or by adding the source repo as a Claude Code plugin marketplace.
How often does this list update?
Quarterly. The search-skill bottom half churns the most; the bioinformatics entries (#9-#11) are stable because the underlying tools (scanpy, ClinVar) change slowly. Expect to see new scientific skills with quality scores climbing into the top 10 as more research-focused authors discover the rubric.
Why is Weather the #1 "scientific & research" skill?
Tag misfit. Weather skills carry weather and forecast tags that get bucketed under science. The skill itself is fine — it's a no-API-key forecast lookup — but it's not what a researcher means by "scientific tooling." The composite score ranks by usage, and a weather skill installed by everyone outranks a bioinformatics skill installed by genomicists. The article exists precisely to surface that distinction.
What's the right starter pack for actual research?
For literature/web research: #3 Prismfy (for arXiv) + #8 YouTube Watcher (for talks). For bench science: install the K-Dense-AI bundle (Biomni #9 + AnnData #10 + ClinVar #11). Three skills covers most research-adjacent workflows.
Related Categories
- Best AI Skills for Data & Analytics in 2026 — analysis side of research overlaps with general data work
- Best AI Skills for Documentation & Writing in 2026 — for the writing-up half of research
- Best AI Skills for AI & Machine Learning in 2026 — for ML researchers specifically
Browse The Full Catalog
Find every skill in this category — including the ones that didn't make the top 15 — at the Scientific & Research page.
Part of the Best-Of 2026 series. Updated 2026-05-18. Skills sampled from a catalog of ~262 active entries with a combined 753.0K popularity signal across the ranked entries.