What is 'hallucination prevention' in this context?

It is a set of rules and agent behaviors that treat all AI output as untrusted until every claim is cross-referenced against a primary source and validated through a verifiable claims table.

How does the /research command work?

The /research command triggers an orchestrated session that decomposes your query, dispatches parallel worker agents to find data, uses verifier agents to check facts, and finally synthesizes everything into a cited report.

How does it rank source quality?

It uses a T1-T6 tiering system where T1 represents primary sources like official documentation or academic papers, while lower tiers include secondary sources and blogs, filtering out SEO spam.

Can I use individual parts of the research methodology?

Yes, you can manually invoke specific principle skills like 'source-evaluation' or 'synthesis-and-reporting' for smaller tasks that don't require a full /research session.

Deep Research Methodology

Name: Deep Research Methodology
Author: oborchers

byoborchers

•

Web Scraping y Recopilación de Datos

Orchestrates multi-agent web research sessions to generate hallucination-resistant reports with verified citations.

This skill transforms ad-hoc web searches into a structured, repeatable research pipeline designed for building high-stakes SaaS products. It leverages a multi-agent architecture where specialized worker, verifier, and synthesizer agents collaborate to decompose complex queries, evaluate source credibility across six tiers, and prevent hallucinations through rigorous citation verification. By automating the transition from information gathering to verifiable reporting, it ensures that every claim is backed by high-quality primary and secondary sources, making it indispensable for technical due diligence and market analysis.

Características Principales

01Multi-agent orchestration using specialized worker, verifier, and synthesizer agents

02Dynamic query decomposition for complex, multi-faceted research topics

03Automated hallucination prevention with circuit breaker patterns and citation checks

04Tiered source evaluation (T1-T6) to prioritize high-quality primary data

05Comprehensive reporting with verifiable claims tables and confidence scoring

069 GitHub stars

Casos de Uso

01Performing market analysis and competitor research for SaaS product development

02Conducting deep technical research on new frameworks or architectural patterns

03Verifying technical claims and gathering primary sources for documentation

Características Principales

01Multi-agent orchestration using specialized worker, verifier, and synthesizer agents

02Dynamic query decomposition for complex, multi-faceted research topics

03Automated hallucination prevention with circuit breaker patterns and citation checks

04Tiered source evaluation (T1-T6) to prioritize high-quality primary data

05Comprehensive reporting with verifiable claims tables and confidence scoring

069 GitHub stars

Casos de Uso

01Performing market analysis and competitor research for SaaS product development

02Conducting deep technical research on new frameworks or architectural patterns

03Verifying technical claims and gathering primary sources for documentation