Deep Research Agent Skills: Academic, Market Intelligence & Multi-Step Research Workflows
Web search returns snippets. Deep research agents return understanding. This category covers skills that go beyond a single query — orchestrating multiple searches, synthesizing sources, reading papers end-to-end, and producing structured reports.
ClawHub hosts deep research skills across 6 distinct workflow types.
Note: Install and download figures in text descriptions reflect stats at the time of writing and may be outdated. All skill tables are live — they fetch current data from the ClawHub database on every page load. Treat table values as authoritative.
Data Overview
| Metric | Value |
|---|---|
| Auto-classified deep research skills | 281 |
| Top by downloads | deep-research-pro ( downloads) |
| Top by installs | market-research ( installs) |
| ArXiv-specific skills | 20+ |
一、General Deep Research Agents
These are orchestration-layer skills: they decompose a research question into sub-queries, run them in parallel or sequence, synthesize results, and return a structured briefing.
deep-research-pro (8.4k downloads) is the flagship — a multi-source research agent that searches the web, synthesizes findings, and produces citation-backed reports. academic-deep-research (6.4k downloads, 54 installs) takes a transparent, methodology-first approach: not a chatbot that sounds authoritative, but a structured research pipeline that shows its work.
gemini-deep-research (5.7k downloads) routes tasks through Gemini's native deep research capability via the gemini-cli MCP extension — for queries that benefit from Gemini's longer context window. parallel-deep-research (2.5k) runs multiple search threads simultaneously via Parallel API, cutting latency for broad research tasks.
| # | Skill | Downloads | Installs | Grade |
|---|---|---|---|---|
| 1 | academic-deep-research | 6,361 | 52 | S |
| 2 | deep-research-pro | 8,370 | 45 | S |
| 3 | gemini-deep-research | 5,693 | 37 | A |
| 4 | parallel-deep-research | 2,469 | 16 | A |
| 5 | academic-research-hub | 1,488 | 10 | A |
| 6 | agent-deep-research | 2,178 | 9 | A |
| 7 | deep-researcher | 639 | 4 | B |
| 8 | cuecue-deep-research | 2,272 | 2 | A |
二、Perplexity-Powered Research
Perplexity's API offers something distinct: AI-generated answers with live citations, not raw search results. These skills integrate Perplexity as a research backend rather than a standalone product.
perplexity (5.6k downloads, 44 installs) is the general-purpose integration. perplexity-deep-search (1.3k downloads) exposes three modes — quick search, deep search, and reasoning — selectable per query. perplexity-safe adds prompt injection guardrails for production deployments.
| # | Skill | Downloads | Installs | Grade |
|---|---|---|---|---|
| 1 | perplexity | 5,623 | 42 | A |
| 2 | perplexity-deep-search | 1,255 | 7 | B |
| 3 | perplexity-search-skill | 1,701 | 2 | B |
| 4 | perplexity-safe | 666 | 0 | B |
三、ArXiv — Academic Paper Research
ArXiv skills represent the richest sub-ecosystem in this category — 20+ skills covering everything from basic paper search to automated daily digests to full pipeline orchestration.
arxiv-watcher (3.8k downloads, 22 installs) is the entry point: search and summarize papers from ArXiv on demand. arxiv (2.9k) covers full-lifecycle access: search, download, and summarize. arxiv-skill-learning (1k) goes a step further — it orchestrates continuous learning by extracting reusable skills from newly published papers.
agentarxiv (3.3k downloads, 10 installs) is the outlier: it's an AI agent that can publish to ArXiv, not just read from it — outcome-driven scientific publishing for AI agents.
For systematic literature reviews, literature-review (4.1k downloads) and literature-search (1.9k) provide structured multi-source review workflows with methodology tracking.
| # | Skill | Downloads | Installs | Grade |
|---|---|---|---|---|
| 1 | arxiv-watcher | 3,830 | 21 | A |
| 2 | arxiv | 2,929 | 21 | A |
| 3 | literature-search | 1,886 | 19 | A |
| 4 | literature-review | 4,081 | 13 | A |
| 5 | arxiv-skill-learning | 1,022 | 11 | A |
| 6 | agentarxiv | 3,270 | 10 | A |
| 7 | arxiv-search-collector | 962 | 10 | A |
| 8 | arxiv-paper-reviews | 1,994 | 9 | B |
| 9 | arxiv-paper-processor | 948 | 8 | A |
| 10 | arxiv-cli-tools | 1,127 | 5 | B |
四、PubMed & Biomedical Research
For biomedical and clinical research, PubMed is the authoritative source — and the NCBI APIs require specialized handling that generic web search doesn't provide.
pubmed-edirect (1.5k downloads) uses NCBI's EDirect command-line interface for structured PubMed queries with field-specific filtering. pubmed2blog converts PubMed papers into patient-friendly, SEO-optimized summaries — useful for medical content teams.
| # | Skill | Downloads | Installs | Grade |
|---|---|---|---|---|
| 1 | pubmed-search-skill | 331 | 4 | B |
| 2 | pubmed-edirect | 1,547 | 2 | A |
| 3 | pubmed2blog | 437 | 1 | B |
五、Market & Competitive Intelligence
Market research skills apply the deep research pattern to business contexts: competitor analysis, market sizing, industry trends, and structured weekly briefings.
market-research (5.6k downloads, 56 installs — the highest install count in this category) is the general-purpose business research agent. competitive-intelligence-market-research (3.6k downloads) covers 24 specific B2B SaaS competitive scenarios across strategy, pricing, and product differentiation.
xiaohongshu-deep-research (3k downloads, 33 installs) is notable for its specificity: a Chinese-market research tool built on academic-deep-research that generates Chinese-language PDF reports — a rare example of a localized fork that materially improves on the original for its target audience.
| # | Skill | Downloads | Installs | Grade |
|---|---|---|---|---|
| 1 | market-research | 5,567 | 52 | A |
| 2 | weekly-report-generator | 2,843 | 36 | A |
| 3 | xiaohongshu-deep-research | 3,019 | 30 | A |
| 4 | market-research-agent | 2,865 | 13 | A |
| 5 | competitive-intelligence-market-research | 3,582 | 7 | A |
六、Fact-Checking & Verification
Verification skills apply structured skepticism to claims: checking news articles, social media posts, and statements against reliable sources with evidence trails.
| # | Skill | Downloads | Installs | Grade |
|---|---|---|---|---|
| 1 | fact-check | 340 | 6 | B |
| 2 | openclaw-fact-checker | 139 | 2 | B |
How to Choose
| I need... | Recommended skill |
|---|---|
| General multi-step research with citations | deep-research-pro |
| Rigorous academic methodology | academic-deep-research |
| AI-answered questions with live sources | perplexity |
| Long-context research (Gemini) | gemini-deep-research |
| ArXiv papers on a topic | arxiv-watcher |
| PubMed / biomedical literature | pubmed-edirect |
| Competitor analysis (B2B SaaS) | competitive-intelligence-market-research |
| Verify a claim or news item | fact-check |
| Publish a finding to ArXiv | agentarxiv |
Last Words
The real gap between search and research is iteration. A single query is search. Research is recognizing that the first answer is incomplete and knowing which follow-up questions to ask — then running them. The skills here are distinguished by whether they implement that loop. deep-research-pro and parallel-deep-research do; most others don't.
ArXiv has become a skill ecosystem of its own. 20+ skills targeting a single API is unusual. It reflects the AI research community's heavy self-use: the people building on ClawHub are also the people reading ArXiv papers, so they've built the integrations they need. The depth here (daily digests, automated paper processing, skill extraction from papers) has no equivalent in any other data source.
market-research has the highest install-to-download ratio in this category. 56 installs vs. 5.6k downloads is 1% — already high for this platform — but for a business-intelligence tool it signals that this is finding regular enterprise use rather than one-time experimentation.
Perplexity vs. deep-research-pro is a depth vs. freshness trade-off. Perplexity answers faster with citations from the live web. Deep research agents go deeper but take longer and may miss breaking news. For time-sensitive questions, Perplexity; for structured analysis, deep-research-pro.
Data source: ClawHub platform download and install stats as of April 16, 2026. Browse more skills at clawhub-skills.com.