CYC2002tommy/Deep-Research-Agent

What it solves

This project provides a specialized research assistant designed to automate the tedious aspects of academic literature reviews. It addresses common AI issues like hallucination and superficial analysis by enforcing a strict, evidence-based pipeline that prioritizes full-text verification over abstract-level summaries.

How it works

It operates as a skill for Claude Code (or compatible runners) using a multi-phase pipeline. It begins with an interview phase to clarify research goals, then executes a structured search strategy across academic databases (like Scopus, Europe PMC, and ScienceDirect). The agent uses a "full-text trigger"—if abstracts are insufficient, it automatically downloads and parses full-text PDFs to extract specific methodologies or data. Every claim is grounded in a verified DOI, and the output undergoes a rigorous "peer-review" gate before final delivery.

Who it’s for

This tool is designed for academic researchers and scientists who need to conduct systematic literature reviews, gather evidence for manuscripts, or build structured knowledge bases while maintaining high standards of academic integrity.

Highlights

• Anti-hallucination design: Every claim is checked against full-text sources, and every DOI is resolved and verified.
• Measured research: Uses a "grill-me" interview method to ensure the agent understands your specific research question before performing any heavy searching.
• Rigorous review: Includes a built-in "Remi" peer-review protocol that acts as a quality gate before any document is finalized.
• Knowledge integration: Automatically generates APA 7th edition reference tables and can sync findings into local Obsidian vaults or RAG systems.
• Flexible search strategies: Offers specialized modes for mapping research fields, identifying knowledge gaps, or creating academic reading lists.