deploy-teamlisted

Deploy a multi-agent team on a hard, multi-dimensional problem where several aspects need simultaneous optimization and trade-offs between them, or to shape the thinking behind a non-obvious architectural choice, or to examine work you've already done from multiple distinct perspectives. This is for throwing compute at genuinely hard problems — not for parallelism. Each agent has its own motivation, runs in an isolated git worktree, coordinates with the others via a shared thread, and is observable through a live dashboard the developer can intervene in. NOT for fire-and-forget parallelism — use the native Agent tool when one specialist can answer.
Rockielab/rockie-claude · ★ 20 · AI & Automation · score 76

Install: claude install-skill Rockielab/rockie-claude

# /deploy-team — multi-agent team on a hard problem The tool is installed at `$HOME/.claude/teams/`. It operates on the repository where the main agent is currently working (resolved via `git rev-parse --show-toplevel`), not where the tool itself is installed. ## When to summon Only deploy a team when ALL THREE hold: 1. **Multi-dimensional problem.** The question has several aspects that must be optimized simultaneously and traded off against each other — not one question with one best answer. 2. **Cross-pollination matters.** You want the agents to read each other's findings mid-work and adjust. If they could produce their answers in isolation, use the Agent tool instead. 3. **Observability is useful.** The developer should be able to watch the work unfold and intervene (post to the thread, DM an agent, pause, stop). If any one is missing, use a specialist subagent via the Agent tool. ## When NOT to summon - One question with one right answer (find all uses of X, fix this bug, audit this function) — use the appropriate subagent - Time-critical work — teams have 30+ second spin-up - Fire-and-forget parallel dispatch — the Agent tool already does that - Anything where you already know the answer and just need it typed — use a specialist ## Flow 1. Compose a team config (JSON). Shape is authoritative at `$HOME/.claude/teams/schema.json` — follow it rather than inventing fields. 2. Show the config to the developer for approval before spending the compute. 3. Invoke from