dsh-agent-teams

When a goal gets complex enough, a single AI session often struggles to play multiple expert roles at once: you constantly switch contexts, manually track task progress, and risk starting downstream work before its dependencies are ready. dsh-agent-teams upgrades the current DeepSeek Harness session into a “captain” that assembles a team of durable sub-agents, splits the goal into dependency-aware tasks, and coordinates the entire execution through direct messages and a live activity dashboard. You just describe the objective in natural language, and the plugin handles team construction and scheduling.

Its core strengths revolve around controlled autonomous collaboration. The captain creates role-specific members backed by continuable sub-agents that can be woken up for focused follow-up turns. Tasks move through explicit states and cannot be claimed before their dependencies finish; idle members automatically pick up the next ready task, while reassignment or captain takeover revokes stale attempts and waits for the old worker to quiesce. Before any real work starts, the plugin stages a roster and task DAG so you can edit member routes, prompts, assignments, and dependencies in the Web UI—only when you hit “Approve & Run” are members spawned and scheduling begun. Optional quality gates enforce requirements → implementation → verification → review → integration contracts, with automatic repair and re-review on failure. Team state is file-backed, so stranded open work survives cold restarts through a retrying scheduler.

This plugin is ideal for developers, tech leads, and product managers who need to turn ambitious goals into multi-step, multi-role collaboration without bolting on an external workflow engine. It gives you both natural-language delegation and a transparent, auditable control panel—the roster, task DAG, member models, and archived history remain visible at all times. For long-running efforts that require handoffs, resumability, or post-hoc review, the persistent members, safe takeover semantics, and complete team archive make it a dependable choice. In short, it shifts you from manually directing AI step-by-step to setting objectives for an AI team and reviewing the final integrated result.