dsh-agent-teams

When a complex goal demands multi-perspective, multi-step collaboration, a single-threaded conversation loses context quickly and sub-tasks queue up serially. dsh-agent-teams turns the current session into a captain: describe the goal in natural language, and the plugin assembles a team of durable sub-agents, splits the goal into dependency-aware tasks, and hands them to a built-in scheduler for parallel execution—no external workflow engine required.\n\nKey capabilities include captain-led delegation with result consolidation, continuable sub-agent members that can be woken for focused follow-up turns, a task state machine that blocks premature claiming until dependencies finish, an idle-member scheduler that atomically claims ready work, durable mailbox messaging between members and the captain, and a live Web activity panel combining segmented progress, a collapsible roster, and an interactive task DAG. All team state is persisted to disk so a scheduler retry after a cold restart picks up stranded work without manual intervention.\n\nBuilt for DeepSeek Harness developers who need multi-agent collaboration: multi-view code reviews, parallel frontend and backend implementation, combined security and performance analysis, or any workflow where one natural-language instruction should spin up a full team and the captain delivers a single consolidated report.