multi-agent-workflows

Orchestrate parallel local agents across Docker, Kubernetes, CI runners, and SSH. Use when decomposing work into independent subtasks, fan-out analysis or refactoring, test sharding, multi-phase pipelines with dependency graphs, or aggregating results from concurrent agents. Triggers include "parallelize this", "run multiple agents", "fan out", "distribute tasks", "orchestrate agents", "run N agents in parallel", "shard tests", or "DAG of agent tasks". Do NOT use for single-shot tasks, `/orchestrate`-suitable jobs, or cloud-managed runs (use `oz agent run-cloud`).