dev-medium

Multi-agent planning is automated by orchestrating architect and planner subagents to investigate problems, generate an INVESTIGATION_REPORT.md, and produce a PLAN.md. You provide the task description, and the workflow automatically stages the problem analysis and plan generation.

Structured investigation and planning is best handled by a coordinated architect-planner workflow that generates explicit artifacts. This approach outputs an INVESTIGATION_REPORT.md for analysis and a PLAN.md to guide execution, ensuring all requirements and frontmatter metadata are explicitly defined.

An architect-planner workflow operates by first deploying an architect subagent to investigate the problem and document findings in INVESTIGATION_REPORT.md. Next, a planner subagent uses this report to formulate a structured execution plan, outputting PLAN.md with defined frontmatter metadata.

Multi-agent orchestration is designed for tasks requiring explicit artifacts, specifically generating INVESTIGATION_REPORT.md and PLAN.md. It applies to medium-complexity problems where structured investigation, staged analysis, and user-facing decision points are necessary for successful execution.

Manual configuration is not required to generate a PLAN.md; you simply provide the task description to start the process. The architect begins the investigation automatically, coordinating with the planner to produce the plan and define the necessary frontmatter metadata.

You should avoid multi-agent planner orchestration for simple tasks that do not require structured investigation or explicit artifacts like INVESTIGATION_REPORT.md. It is specifically built for medium-complexity problems needing staged analysis, automated plan generation, and user-facing decision points.