Orchestrate a multi-agent implementation: decompose a task, fan out subagents to implement each piece in parallel, and brief each one with the context it needs plus a mandate to pull fresh docs via Context7. Use when the user wants to "use subagents", "fan out", "parallelize implementation", "orchestrate agents", spin up workers, or build a feature that splits into independent workstreams. Covers Claude Code subagents, Claude Code dynamic workflows, and Codex subagents.
