spawn

---
name: spawn
description: Use when launching N parallel subagents in isolated git worktrees to work on a session task; each agent explores a different strategy and commits results.
category: Engineering
version: 1.0.0
tools: []
---

# /hub:spawn — Launch Parallel Agents

Spawn N subagents in isolated git worktrees to work on the same task in parallel, each exploring a different strategy.

## Usage

```
/hub:spawn                                    # Spawn agents for latest session
/hub:spawn 20260317-143022                    # Spawn for specific session
/hub:spawn --template optimizer               # Use optimizer template
/hub:spawn --template refactorer              # Use refactorer template
```

## Templates

When `--template <name>` is provided, use the dispatch prompt from agent-templates.md:

| Template | Pattern | Use Case |
|----------|---------|----------|
| `optimizer` | Edit → eval → keep/discard → repeat x10 | Performance, latency, size reduction |
| `refactorer` | Restructure → test → iterate until green | Code quality, tech debt |
| `test-writer` | Write tests → measure coverage → repeat | Test coverage gaps |
| `bug-fixer` | Reproduce → diagnose → fix → verify | Bug fix with competing approaches |

Replace all `{variables}` with values from session config. Assign each agent a **different strategy** — diverse strategies maximize parallel exploration value.

## What It Does

1. Load session config from `.agenthub/sessions/{session-id}/config.yaml`
2. For each agent 1..N:
   - Write task assignment to `.agenthub/board/dispatch/`
   - Build agent prompt with task, constraints, and board write instructions
3. Launch ALL agents in a **single message** with multiple Agent tool calls (true parallelism)
4. Each agent:
   - Works in its worktree (isolated filesystem)
   - Commits changes with descriptive messages
   - Writes result summary to `.agenthub/board/results/agent-{i}-result.md`
   - Includes: approach taken, files changed, metric, confidence level
5. Update session state to `running`

## Critical Rules

- **All agents in ONE message** — spawn all Agent tool calls simultaneously
- **isolation: "worktree"** mandatory — each agent needs isolated filesystem
- **Never modify session config** after spawn — agents rely on stable config
- **Each agent gets a unique board post** — dispatch posts numbered sequentially
- Agents do NOT read or modify other agents' work

## After Spawn

Tell the user:
- {N} agents launched in parallel
- Each working in an isolated worktree
- Monitor with `/hub:status`
- Evaluate when done with `/hub:eval`