Deep Research

Autonomous multi-step research that searches multiple sources, reads full content, synthesizes findings, and produces a structured report.

When to Use

• User wants a thorough understanding of a topic (medical condition, drug, treatment, technology)
• User asks for a literature review or evidence summary
• User wants competitive or landscape analysis
• User wants to investigate an open question with multiple angles
• User asks "what does the research say about X"

Research Strategy

Step 1: Query Decomposition

Break the research question into 3–5 sub-questions covering:

• Core definition / mechanism
• Current evidence / state of the art
• Debates, limitations, or contradictions
• Clinical / practical implications (if medical)
• Recent developments (last 1–2 years)

Step 2: Multi-Source Search

Run searches across complementary sources using the available search tools:

# Use multi-search-engine for broad web coverage
# Use pubmed-search for peer-reviewed medical literature
# Use agent-browser to read full-text articles and retrieve content blocked by snippets

Search order:

1. PubMed (if medical/biomedical topic) — for peer-reviewed evidence
2. Multi-search-engine (Bing, Google, DuckDuckGo) — for guidelines, reviews, news
3. Wikipedia — for background and structured overviews
4. agent-browser — for reading full articles, PDFs, clinical guidelines

Step 3: Source Evaluation

For each source note:

• Publication type (RCT, meta-analysis, guideline, review, news)
• Date (prefer sources within 5 years for medical topics)
• Authority (journal impact, organization credibility)
• Relevance to the specific sub-question

Step 4: Synthesis

Synthesize across sources into a coherent narrative. Do NOT just concatenate summaries — identify: