dev-manager

Multi-agent coordination for a shared story is achieved by decomposing work into non-overlapping linked tasks with explicit ownership. A dev-manager orchestrates subagents and human worktrees under a single contract to prevent drift and ensure a unified outcome.

Task decomposition for parallel execution involves splitting large problems into non-overlapping units with a shared seam for interlock. This establishes explicit ownership boundaries and decision points so multiple agents can progress without colliding.

Yes, multi-agent orchestration applies to both automated subagents and human-driven worktrees. It coordinates linked tasks under a shared objective, monitoring progress across both types of contributors and intervening only on collision or drift.

Merge workflows handle collisions by enforcing continuous monitoring and intervention when drift or collisions are detected. The orchestration layer applies secure, auditable merges to ensure all parallel work integrates cleanly toward the story goal.

Use a multi-agent dev manager for complex development efforts requiring coordinated parallel work rather than a single agent. It is necessary when tasks must be decomposed into non-overlapping units with strict ownership boundaries and auditable merge workflows.

To start orchestrating dev agents, provide the story objective. The dev-manager will spawn subagents, decompose the problem into linked tasks, establish guardrails, and begin coordinating the parallel work automatically.