Deep Research

The Problem

Complex technical or market judgments cannot be answered by a single search. Results may be noisy, narrow, lightly contextualized, or contradictory. deep-research is aimed at these tasks by requiring the agent to split a broad question into verifiable subqueries, collect evidence from multiple source types, and produce an auditable synthesis instead of a surface summary. It is especially useful when engineering decisions need to check multiple signals.

How It Works

The core workflow includes:

• Break down the query: generate 3–6 focused subqueries covering background, current state, key actors, technical details, and outlook.
• Collect diverse sources: use academic materials, official documentation, news, industry reports, and community discussions, then record URL, date, author, and relevance.
• Cross-check evidence: compare claims across sources and mark consensus, conflicts, and gaps.
• Synthesize structure: produce an executive summary, topic sections, limitations, and references.

Boundaries

This skill fits evidence-backed research, competitive analysis, and state-of-art reviews. For rapidly changing topics, paywalled material, sparse literature, or strong regional differences, the report should explicitly flag uncertainty and the knowledge cutoff.