Conduct In-depth Research and Generate Comprehensive Reports

Research Agent autonomously investigates a given topic and produces a comprehensive report. It follows a four-step process: understanding the question (clarifying the research objective, identifying key terms and concepts, defining scope and boundaries), gathering information (searching the codebase for relevant implementations, reading documentation and comments, using web search for external context when enabled, and cross-referencing multiple sources), analyzing findings (identifying patterns and connections, noting contradictions or gaps, evaluating source reliability, considering multiple perspectives), and synthesizing a report (summarizing key findings, providing evidence and citations, highlighting uncertainties, suggesting follow-up questions).

Use it to investigate an open-ended topic - inside a codebase, in documentation, or on the web when enabled - and get a structured, cited report rather than a raw answer. If the research scope is unclear, the agent is designed to ask clarifying questions before starting rather than guessing at intent.

Input is a research topic, optionally with a defined scope. Output is a structured research report: a 2-3 sentence summary, a numbered list of key findings each backed by evidence, a list of sources consulted, an explicit confidence level (high, medium, or low), and recommended next steps as concrete action items.

### Research Report: [Topic]

**Summary**: 2-3 sentence overview

**Key Findings**:
1. Finding with evidence
2. Finding with evidence
3. Finding with evidence

**Sources Consulted**:
- [source 1]
- [source 2]

**Confidence Level**: High/Medium/Low

**Recommended Next Steps**:
- Action item 1
- Action item 2

Anyone who needs a thorough, evidence-backed answer to an open-ended question - codebase archaeology, documentation review, or web research - delivered as a structured report with an explicit confidence level rather than an unqualified answer.