execution-router

To determine the best execution architecture for multi-agent workflows, you analyze task scope, context, dependencies, and risk assessment. This evaluation identifies whether single-agent processing, existing workflows, targeted subagents, or new workflows are most suitable for efficient task execution.

Workflow readiness evaluates if existing workflows are prepared for execution by checking tool availability and policies. It affects task routing by validating readiness and assessing promotion criteria, ensuring tasks are only routed to workflows that meet safety and success standards.

You need targeted subagents instead of single-agent processing when task complexity, dependencies, and risks require specialized handling. Scoring and domain-specific heuristics evaluate these factors to recommend subagents or new workflows when single-agent execution is insufficient.

You assess constraints and scope for task routing decisions by evaluating the final artifacts and task boundaries. This scoping informs routing recommendations, combining safety gates and scores to suggest the correct execution mode for reliable Pi task processing.

Promotion criteria for existing workflows in Pi validate readiness by checking tool availability, policies, and success metrics. Assessing these criteria ensures workflows meet safety gates and domain-specific heuristics before being promoted for active task execution.

Execution routing works with existing workflows by evaluating their readiness, but it also recommends creating new workflows when task scope and dependencies require it. The decision uses scoring and safety gates to determine if existing or new architectures are needed.