run

You can execute a complete multi-agent workflow by chaining initialization, baseline capture, agent spawning, evaluation, and merge confirmation in a single guided flow. This orchestrates the full agent lifecycle in one shot.

The best way to compare competing agent outputs is to spawn multiple agents at once, monitor their progress, and evaluate the results using metric-based baselines or LLM judge mode when no metric command is provided.

To ensure safe merging during parallel agent execution, the workflow presents the winning result and waits for your explicit user approval before merging any changes into the codebase.

Yes, you can use workflow orchestration for test generation and comparative prompt-based tasks. It applies to multi-agent software engineering tasks including optimization, refactoring, bug fixing, and benchmarking.

You do not need to provide a metric command for agent evaluation. The workflow supports metric-based evaluation with baselines, but automatically switches to LLM judge mode when no metric command is provided.

Your multi-agent task requires sequential execution phases because the lifecycle orchestrates initialization, baseline capture, agent spawning, evaluation, and merge confirmation in a strict order to ensure valid comparative results.