This context describes a local research agent that investigates questions against public web sources and produces evidence-backed reports.
Research Question: The user's question that defines the subject and intended outcome of one research run. Avoid: Prompt, query
Research Plan: The agent's evolving decomposition of a Research Question into focused investigations. Avoid: Task list, chain of thought
Source: A publicly accessible web resource used as evidence during a research run. Avoid: Document, context
Evidence: Information extracted from a Source and retained with enough provenance to support or challenge a claim. Avoid: Fact, snippet
Research Report: The Markdown deliverable that answers a Research Question and connects key claims to Sources. Avoid: Response, summary
Research Budget: The configured limits on searches, Source reads, and elapsed time for one research run. Avoid: Token budget, quota
Research Depth:
The user-facing quick, standard, or deep preset that selects a Research Budget while allowing the agent to stop early when the Research Question is sufficiently answered.
Avoid: Search count, iteration count