agent-orchestrator

You orchestrate sub-agents by providing a macro task to an orchestrator that automatically decomposes it into independent subtasks, spawns specialized agents, and consolidates results. It uses file-based communication with inbox/outbox directories to manage state and parallel execution.

File-based communication for AI coordination uses generated SKILL.md files, inbox/outbox directories, and status.json tracking to manage sub-agent state and results. This allows autonomous agents to exchange information and track progress without real-time network dependencies.

Yes, coordinating multi-agent workflows requires a Task tool to spawn sub-agents and provision workspaces. The orchestrator depends on this tool to generate SKILL.md files, manage inbox/outbox communication, and track status.json for each autonomous sub-agent.

Yes, sub-agent orchestration applies to complex, multi-step projects across software, research, and operations where tasks can be parallelized. The orchestrator decomposes macro tasks into independent subtasks with clear success criteria for specialized agents.

To start automating multi-step projects, provide a macro task to the orchestrator. It will decompose the task, spawn specialized sub-agents with provisioned workspaces, monitor their progress, and produce a final consolidated result upon completion.

File-based communication for workflow automation requires generating SKILL.md files and status.json tracking for each sub-agent. It relies on a Task tool to provision workspaces and manage inbox/outbox directories, making it dependent on structured file state management rather than dynamic memory.