Two-Agent System — Specialist and Reviewer Agent Pair

I will build a two-agent system using a specialist-reviewer pattern — one agent performing the primary task (research, writing, analysis, or code generation) and a second agent reviewing the output against defined quality criteria and requesting revisions until quality standards are met. The specialist-reviewer pattern is the simplest multi-agent architecture that delivers measurable quality improvement over a single agent — the reviewer agent consistently catches errors, omissions, and quality issues that the specialist agent's self-review misses, because the reviewer role activates a different reasoning mode than the production role.

The system covers specialist agent configuration with task-focused system prompt and tools, reviewer agent configuration with quality criteria and critique guidelines, an iterative revision loop with configurable maximum iterations, and a final output gate requiring reviewer approval.