skills.rest
Oatse
OatseCOMMUNITY
Explore all skills by this author →
GitHub

deep-research

Conducts parallel web research across multiple sources and produces cited Markdown reports with confidence tracking.

Updated May 23, 2026
Search Tags:#web search#report generation#citations#market analysis#literature review#competitive analysis#deep research
One-click install
npx skills add https://github.com/Oatse/CWE-Automation --skill deep-research-oatse
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: deep-research
Source: https://github.com/Oatse/CWE-Automation/tree/main/.agents/skills/deep-research
Command: npx skills add https://github.com/Oatse/CWE-Automation --skill deep-research-oatse

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires pandoc, md-to-pdf, and includes references (resource) and assets (resource) components.

What problem does it solve? Producing a thorough, well-sourced research report requires scanning dozens of sources, reconciling conflicting figures, and tracking which claims are verified — work that is slow, error-prone, and often results in unsourced assertions presented as fact. ## Core Features & Use Cases - Parallel multi-source research: Fans out 3–20 sub-agents across independent research axes, each searching the web and returning findings with inline citations, access dates, and confidence levels. - 11 research types: Predefined axis breakdowns and section templates for market, domain, technical, competitive, product, academic, person/org, financial, legal, trend, and community research. - Citation and confidence discipline: Every claim carries a source URL, accessed date, and High/Medium/Low confidence level; critical claims require 2+ independent sources and conflicts between sources are surfaced explicitly. - Use Case: Ask for a competitive analysis of three SaaS products and receive a structured Markdown report with competitor profiles, pricing comparison tables, review sentiment, and win/loss signals — every figure cited and confidence-rated. ## Quick Start Ask the agent to research a topic, for example: research the global cybersecurity market size, growth projections, and top demand drivers, and produce a cited report.

Frequently Asked Questions about deep-research

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

FAQPage Schema
How do I run a deep research report on a market or topic?▼

Ask the agent to research your topic, for example "research the plant-based food market size and key players". The skill scopes the request, fans out parallel web-search sub-agents across research axes, and writes a cited Markdown report to ./research/ with findings appended as each agent completes.

What research types does the deep research skill support?▼

It supports 11 types: market, domain, technical, competitive, product, academic, person/org, financial, legal, trend, and community. Each type has a dedicated reference file defining research axes, probing questions, and a report section template.

How are sources and confidence levels handled in research reports?▼

Every claim gets an inline citation with source URL, accessed date, and a High, Medium, or Low confidence level. Critical claims like market size require two or more independent sources, and conflicting figures are surfaced explicitly rather than silently resolved.

Does the deep research skill require internet access?▼

Yes, WebSearch and WebFetch are mandatory; the skill halts immediately if web search is unavailable. It is designed for Claude Code or similar agents with internet access, and curl is used for downloading external documents like PDFs.

Can I export the research report to PDF?▼

Yes, after the Markdown report is finalized the skill offers an optional PDF export step. It tries pandoc first with wkhtmltopdf, weasyprint, or a LaTeX engine, then falls back to the md-to-pdf Node tool if pandoc is unavailable.

What happens when sources disagree on a figure?▼

The skill never silently picks one figure. It reports the conflict with both sources cited, explains likely causes such as methodology differences, and either uses a conservative baseline with explicit hedging or flags the claim as Low confidence.

Related Skills

Explore More Education & Research Skills→
danny0926danny0926

research-scope

Define research scope with hypotheses, bias checks, and prior art references.

Community
Advanced
tomaszstaniaktomaszstaniak

para-method

Organizes captured knowledge using PARA decision framework and CODE workflow for reuse.

Community
Intermediate
zero3041zero3041
1

data-scraper-agent

Automate public data collection, AI enrichment, and database storage.

Community
Advanced
plurigridplurigrid
60

depth-search

Query local ~/.topos notes and academic MCPs in parallel for multi-source research.

Official
Advanced