skills.rest
hint-shu
hint-shuCOMMUNITY
Explore all skills by this author →
GitHub

research

Decomposes queries into sub-questions and synthesizes cited reports from scraped web sources.

4|Updated Apr 17, 2026
Search Tags:#research#report generation#firecrawl#citations#web scraping#tavily#query decomposition
One-click install
npx skills add https://github.com/hint-shu/deep-research --skill research-hint-shu
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: research
Source: https://github.com/hint-shu/deep-research/tree/main/skills/research
Command: npx skills add https://github.com/hint-shu/deep-research --skill research-hint-shu

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Ad-hoc web searches produce shallow answers with one or two sources and no audit trail. This Skill runs a structured L1 research pipeline that decomposes a query, scrapes 10-15 sources, writes per-source summaries, and synthesizes a cited ~1000-word report. ## Core Features & Use Cases - Planner Decomposition: Breaks any research question into exactly 3 sub-questions saved to plan.md before searching. - Parallel Search & Scraping: Runs Firecrawl and Tavily searches concurrently, then scrapes and deduplicates 10-15 ranked sources. - Per-Source Summaries: Writes a 300-500 word .sum.md file per source with quality grades, key facts, and quotes, enabling traceable synthesis. - Verified Output: Runs a shared verification script checking word count, citation mapping, and source pairs before delivery. - Use Case: Ask "how does Drizzle ORM work" and receive a structured report with inline citations, an annotated bibliography, and all artifacts stored in .firecrawl/research/<slug>/L1/. ## Quick Start Ask the assistant to research how a specific framework or technology works using the research skill.

Frequently Asked Questions about research

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a structured web research workflow with citations?▼

Invoke the research skill with a clear question. It decomposes the query into 3 sub-questions, searches with Firecrawl and Tavily in parallel, scrapes 10-15 sources, and produces a report where every factual claim has an inline [N] citation mapped to a bibliography.

What is the difference between quick-research and deep-research tiers?▼

Quick-research (L0) gives a one-paragraph answer from 3-5 sources in about a minute. This L1 research tier adds planner decomposition and per-source summaries for a ~1000-word report, while deep-research (L2) adds reflection and contradiction checking on top of L1 output.

Does the research skill work with Russian-language queries?▼

Yes, the skill matches the user's query language in output and supports bilingual search strategies. For Russian-language topics it runs one query in Russian and one in English to broaden source coverage.

When should I not use the L1 research tier?▼

Skip L1 for simple fact-checks, where quick-research suffices, and for comparisons or strategic decisions, where deep-research or expert-research are designed. L1 targets non-trivial but non-strategic overview questions.

Why does the research report verification fail?▼

Verification fails if the report is under 700 words, has fewer than 10 scrape-summary pairs, or contains citations not mapped to bibliography entries. It also fails if the shared verify-research.sh library is missing, requiring scripts/install.sh to be run first.

Related Skills

Explore More Education & Research Skills→
ZJU-REALZJU-REAL
195

firecrawl

Search, scrape, and monitor web content using the Firecrawl CLI.

Official
Advanced
Vibecoder79Vibecoder79

research

Routes research questions between Claude WebSearch and Perplexity sonar-deep-research via OpenRouter.

Official
Intermediate
julianobarbosajulianobarbosa

OSINT

Conduct structured OSINT investigations across people, companies, domains, and organizations.

Community
Advanced
oneryalcinoneryalcin

Web Research

Automate web research with source verification, credibility tagging, and confidence scoring.

Community
Basic