Best Claude Skills for Customer Research in 2026: 15 Compared
Top 15 Claude skills for research work. Academic-paper skills lead by signal; last30days (#8, 5/5 quality) is the standout for actual customer/market research across Reddit, X, HN, Polymarket.
Customer Research is the category where 2026's "research" definition has split. The top-of-list skills (#1-#4) are about academic research — paper writing, paper reproduction, citation hunting. Customer/market research as a discipline shows up at #8 (last30days, 5/5 quality) which scans Reddit/X/HN/YouTube/Polymarket from the past 30 days. The middle of the list is dominated by web search skills (#5-#7, #9, #11, #12) — useful for research broadly but generic. The honest take: pick by which kind of research you do. Academic researchers want #1-#4; product/marketing researchers want #8.
Quick Pick
last30days — 5/5 quality, 20K signal. Scans Reddit, X, HN, YouTube, Polymarket from the past 30 days, synthesizes findings into grounded summaries with citations. The single most useful customer/market research skill in the catalog. (Note: pick something else if you do academic research — see #1-#4.)
What These Skills Actually Do
Research skills split into three patterns in 2026: (1) Academic research (#1 research-paper-writing, #4 paper reproduction, #13 Academic Deep Research) — for ML/scientific paper writers. (2) Customer/market research (#8 last30days) — scan trend signals across consumer platforms, output cited summaries. (3) Generic web search (#5 Multi Search Engine, #6 Prismfy, #7 #11 Tavily variants, #9 Brave, #12 DuckDuckGo wrapper) — useful for any research that's really "search the web and synthesize." The fourth honorary pattern is narrow-focus data feeds (#3 Polymarket via Hermes for prediction markets, #15 Financial Search for finance, #10 YouTube transcripts). The 2026 lesson: customer research is no longer just "talk to customers" — it's "talk to customers plus systematically scan Reddit/HN/X for what they're already saying." The skills that do the latter (#8 specifically) are massively underrated.
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. research-paper-writing (Hermes)
Skill · NousResearch/hermes-agent · 124.8K signal · quality 4/5 Write ML papers for NeurIPS/ICML/ICLR.
The take: Academic ML research workflow. The conference-specific framing (NeurIPS, ICML, ICLR) makes it useful for AI researchers; useless for product/market research.
2. blogwatcher (Hermes)
Skill · NousResearch/hermes-agent · 124.8K signal · quality 4/5 RSS/Atom feed monitoring via blogwatcher-cli.
The take: Diversification cap explains the same 124.8K signal as #1 and #3 — all three are from the same Hermes repo. Useful for monitoring industry blogs at scale.
3. polymarket (Hermes)
Skill · NousResearch/hermes-agent · 124.8K signal · quality 4/5 Polymarket queries — markets, prices, orderbooks, history.
The take: Prediction markets as a research signal. Real money attached, real honesty. Install for macro/political/tech forecasting research.
4. AI Paper Reproduction Workflow
Plugin · lllllllama/ai-paper-reproduction-skill · 49.1K signal · quality 4/5 4-skill workflow for reproducing AI papers — context, env bootstrap, repo intake, audit.
The take: Solves the "I read this paper but can't get the code running" problem. Strong workflow scaffolding. Niche to ML researchers but invaluable in that niche.
5. Multi Search Engine
Skill · clawhub/multi-search-engine · 102.0K signal · quality unrated 17 search engines (8 CN + 9 Global), no API keys.
The take: Cross-listed from Scientific & Research. Most comprehensive search wrapper; install once and skip the other search skills here.
6. Prismfy Web Search
Skill · uroboros1205/prismfy-search · 77.5K signal · quality unrated 10 engines including arXiv, Reddit, GitHub, HN.
The take: Differentiated by arXiv and HN inclusion — these matter for technical/founder research more than generic web search.
7. Tavily 搜索
Skill · jacky1n7/openclaw-tavily-search · 70.8K signal · quality unrated Tavily API — LLM-optimized.
The take: First of three Tavily wrappers (also #11). Picks one Tavily skill; runs three burns context unnecessarily.
8. last30days
Skill · mvanhorn/last30days-skill · 20.2K signal · quality 5/5 Scans Reddit, X, HN, YouTube, Polymarket from past 30 days. Citation-grounded summaries.
The take: The standout skill on this list. 5/5 quality. The "30-day window" constraint is the value — by-design recency forces grounded analysis instead of training-cutoff guessing. v2.9.5 with 455 tests shows the author cares about quality. Install this for any customer/market research.
9. Brave Search
Skill · steipete/brave-search · 47.6K signal · quality unrated Brave API — lightweight, no browser.
The take: Privacy-focused search. Paid API key, fast responses. Alternative to Tavily; pick one.
10. YouTube Watcher
Skill · michaelgathara/youtube-watcher · 39.0K signal · quality unrated YouTube transcripts — summarize, Q&A, extract.
The take: Cross-listed. For customer research specifically, useful for analyzing competitor demos, founder podcasts, and longform interview content.
11. Tavily AI Search
Skill · bert-builder/tavily · 27.7K signal · quality unrated Tavily wrapper with AI-summarized answers.
The take: Third Tavily variant. Pick one. The AI-answer-summary feature is the differentiator over #7.
12. Web Search (DuckDuckGo)
Skill · billyutw/web-search · 27.2K signal · quality unrated DuckDuckGo search wrapper.
The take: Free, no API key. Lighter than the paid alternatives (#7, #9, #11). Quality of results is lower; quality of privacy is higher.
13. Academic Deep Research
Skill · kesslerio/academic-deep-research · 15.5K signal · quality unrated Transparent rigorous research — 2-cycle research per theme, APA 7th, evidence hierarchy.
The take: The "not a black-box API wrapper" framing is the value — most research skills hide methodology, this one exposes it. Useful for academic and consulting deliverables where the methodology matters as much as the conclusion.
14. Cellcog
Skill · nitishgargiitd/cellcog · 12.6K signal · quality unrated Any-to-any AI sub-agent.
The take: Cross-listed (#8 in Video Production). Same caveat — too broad to be authoritative on research specifically.
15. Financial Search Engine
Skill · financial-ai-analyst/mx-finance-search · 11.2K signal · quality unrated NL search for financial info — announcements, reports, news, policies. Global market coverage.
The take: Finance-vertical version of #5/#6. Install for investment research; skip otherwise. Chinese-keyword triggers work better.
Comparison Table
| # | Skill | Type | Stars / Installs | Quality | License |
|---|---|---|---|---|---|
| 1 | research-paper-writing | Skill | 124.8K | 4/5 | MIT |
| 2 | blogwatcher | Skill | 124.8K | 4/5 | MIT |
| 3 | polymarket | Skill | 124.8K | 4/5 | MIT |
| 4 | AI Paper Reproduction Workflow | Plugin | 49.1K | 4/5 | MIT |
| 5 | Multi Search Engine | Skill | 102.0K | — | MIT-0 (Free to use, modify, and redistribute. No a |
| 6 | Prismfy Web Search | Skill | 77.5K | — | — |
| 7 | Tavily 搜索 | Skill | 70.8K | — | — |
| 8 | last30days | Skill | 20.2K | 5/5 | — |
| 9 | Brave Search | Skill | 47.6K | — | — |
| 10 | YouTube Watcher | Skill | 39.0K | — | — |
| 11 | Tavily AI Search | Skill | 27.7K | — | — |
| 12 | Web Search | Skill | 27.2K | — | — |
| 13 | Academic Deep Research | Skill | 15.5K | — | — |
| 14 | Cellcog | Skill | 12.6K | — | — |
| 15 | Financial Search Engine | Skill | 11.2K | — | — |
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?
Skills are open source. Several depend on paid API keys: Tavily (#7, #11), Brave (#9), Polymarket API access. The standout #8 last30days uses Claude's own access patterns — no extra 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. Research skills are stable; the #8 standout is at v2.9 which suggests active maintenance.
What if I do customer interviews, not desk research?
This list is desk-research only — scanning signals, synthesizing content. For actual interview workflows, use the Elicitation skill from the Specialized Roles list, which encodes proper research-interview methodology.
What's the right starter pack for customer research?
Two skills: #8 last30days (the discipline-defining install) + one web search wrapper of your choice (#5 if you want maximum coverage, #9 if you want speed). Add #10 YouTube Watcher for longform competitor analysis.
Related Categories
- Best AI Skills for Scientific & Research in 2026 — broader research category
- Best AI Skills for Business & Marketing in 2026 — market research overlaps
- Best AI Skills for Specialized Roles in 2026 — Elicitation for interview workflows
Browse The Full Catalog
Find every skill in this category — including the ones that didn't make the top 15 — at the main browse page.
Part of the Best-Of 2026 series. Updated 2026-05-18. Skills sampled from a catalog of ~262 active entries with a combined 874.9K popularity signal across the ranked entries.