dsh-agent-teams

One prompt, a working team: turn the current DSH session into a captain that assembles durable sub-agents, splits a goal into dependency-aware tasks, and coordinates over direct messages.

What problem it solves

Complex multi-actor work is hard to decompose inside a single session: who does what, what must finish first, and how results come together usually require ad-hoc assembly or manual coordination. This plugin turns the current DeepSeek Harness session into a captain that assembles durable sub-agents, splits a goal into dependency-aware tasks, and coordinates work through direct messages — no separate workflow engine required.

It ships a team protocol, ten coordination tools, persistent state, an automatic shared-task scheduler, and a live Web UI with an interactive task DAG. Members are continuable DSH sub-agents: idle members claim the next ready task, reassignment revokes stale attempts before new work starts, and cold recovery retries stranded open attempts. The captain consolidates the final result and archives the complete team record.

Key features

• Captain-led delegation: the current session creates the team, assigns roles, and consolidates the final result
• Durable members: members are continuable DSH sub-agents that can be woken for focused follow-up turns
• Dependency-aware tasks: tasks move through explicit states and can’t be claimed before their dependencies finish
• Automatic reuse and safe takeover: idle members claim the next ready task; reassignment revokes stale attempts, and cold recovery retries stranded open ones
• Direct messaging: members send durable mailbox messages straight to teammates or the captain, no relay
• Live activity panel: segmented progress, a collapsible roster, and an interactive task DAG; completed archives retain full member and task history
• Slash command: the /agent-teams host command activates the protocol without fixed phrasing

Quick start

Install:

dsh plugin --profile web add @nanmicoder/dsh-agent-teams

Restart DSH and refresh the Web UI, then describe the goal directly:

Use AgentTeams to review the commits after v0.5.3 from performance, security, and product perspectives. Return one consolidated report.

You can also pick agent-teams from the Web GUI slash menu, type the goal, and press Enter. Defaults work out of the box; a trusted profile can override stateDir, memberProvider (spawn / fork), maxMembers, and more. Team state lives under <workspace>/.agent-teams/.

Who it is for

• DSH users whose tasks are complex enough to split across several sub-agents working in parallel
• Anyone who wants structured multi-agent coordination (roles, dependencies, progress, reporting) instead of an ad-hoc assembly
• Teams that need an auditable record of members, tasks, and archived history