blueprint

To create a multi-agent plan, you provide a single goal that gets transformed into a step-by-step blueprint. This decomposition includes per-step context summaries, dependency graphs, and evaluation criteria so new agents can execute tasks directly without losing context.

Multi-agent orchestration coordinates multiple agents across complex tasks by generating a structured workflow. It detects parallel steps, tracks dependencies, and implements stage reviews to ensure safe concurrent execution across migrations or large feature rewrites.

Managing cross-agent dependencies is handled by mapping a goal into a structured blueprint with explicit dependency graphs. The workflow ensures tasks requiring parallel work are properly sequenced and includes rollback strategies to prevent execution failures.

Yes, this is designed specifically for coordinating multiple agents across PRs, migrations, and large feature rewrites. It breaks down the overarching goal into trackable steps with per-step context, ensuring seamless handoffs between concurrent sessions.

Multi-agent workflows are protected by anti-pattern directories, adversarial review gates, and mutation protocols. These mechanisms ensure plan quality and safety by preventing unsafe parallel execution and providing rollback strategies during stage reviews.

You should avoid structured multi-agent blueprints for simple, single-session tasks that lack cross-agent dependencies. The overhead of dependency tracking, parallelism checks, and mutation protocols is unnecessary without complex parallel work or large feature rewrites.