Orchestrator
Runs multiple agents in parallel waves.
What it does
/orchestrator breaks a complex task into independent subtasks and runs them in parallel using multiple Claude Code agents. Instead of one agent doing everything sequentially, you get several agents working simultaneously, each on a focused piece, with the results synthesized at the end.
The speed improvement on multi-part tasks is significant. A task that takes 30 minutes in sequence might take 8 minutes in parallel.
When to use it
- When a task has multiple independent parts that don't need each other to complete
- Research tasks where you need to gather information from several sources simultaneously
- Review tasks where different agents can check different aspects (security, style, logic) at the same time
How to trigger it
Describe the task and let the skill break it down: orchestrate a review of this PR
Or specify the waves: run these three research tasks in parallel: [list]
What you need
Claude Code CLI with enough API capacity for parallel agent runs. The orchestrator creates sub-agents in your Claude Code session. Works best for tasks where the subtasks are genuinely independent. If subtask B depends on the output of subtask A, sequential is the right approach.
Ready to install?
PM Pilot lives on GitHub. Every skill is a plain markdown file you can read, edit, and install in under 5 minutes.