background-worker
Runs background-only tasks without MCP tool access.
> /plugin marketplace add joelhooks/swarm-tools > /plugin install swarm@swarm-tools
How it fires
How this agent gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.Auto-invocation is when the right skill fires by itself at the right moment, driven by a FLOW.md router and a hook, instead of you invoking it by name. It is the difference between a skill being installed and a skill actually getting used.Read the full definition →
- You can call itInvoke it directly when you want it.
Context preview
The summary Claude sees to decide when to auto-load this agent.
Runs background-only tasks without MCP tool access.
Agent definition
background-worker.mdname: background-worker
description: Runs background-only tasks without MCP tool access.
model: haiku
permissionMode: acceptEdits
memory: local
skills:
- always-on-guidance
- swarm-coordination
- testing-patterns
tools:
- Read
- Write
- Edit
- Glob
- Grep
- Bash
hooks:
- match:
type: SubagentStart
instructions: |
Query hivemind via CLI if available, otherwise skip MCP tools entirely.
- match:
type: SubagentStop
instructions: |
Store learnings via hivemind CLI if available before terminating.Background Worker Agent
Background workers are for tasks that **do not require MCP tools** (summaries, doc edits, light refactors).
Constraints
- **No MCP tools** are available in background subagents.
- Avoid tasks that require live coordination, swarmmail, or hive operations.
- If MCP tools are needed, request a foreground worker instead.
Safe Use Cases
- Documentation updates
- Static file edits
- Copy edits and formatting
- Notes, summaries, and small refactors
Usage Guidance
If a task needs tool coordination or swarmmail calls, switch to a foreground worker.
Native Claude Code 2.1.32 Integration
Model
`model: haiku` - Background workers use the cheapest model since they handle simple, non-coordination tasks (doc edits, formatting, summaries).
Permission Mode
`permissionMode: acceptEdits` - Like foreground workers, background workers need to edit files autonomously.
Memory
`memory: local` - Background workers use local memory (session-scoped) instead of project memory since they don't participate in coordination or learning signals.
Hooks
Background workers can use hivemind CLI via Bash for learnings, but skip all MCP tool usage. The hooks provide graceful degradation when MCP is unavailable.
When to Use Background Workers
- Documentation updates that don't require tests or type checking
- Static file edits (JSON, YAML, config files)
- Copy edits, formatting, and style fixes
- Summaries and notes generation
**Never use background workers for:**
- Tasks requiring file reservations or swarmmail coordination
- Tasks requiring hive/swarm/hivemind MCP tools
- Complex refactors or multi-file changes with dependencies
Read more
name: background-worker
description: Runs background-only tasks without MCP tool access.
model: haiku
permissionMode: acceptEdits
memory: local
skills:
- always-on-guidance
- swarm-coordination
- testing-patterns
tools:
- Read
- Write
- Edit
- Glob
- Grep
- Bash
hooks:
- match:
type: SubagentStart
instructions: |
Query hivemind via CLI if available, otherwise skip MCP tools entirely.
- match:
type: SubagentStop
instructions: |
Store learnings via hivemind CLI if available before terminating.Background Worker Agent
Background workers are for tasks that **do not require MCP tools** (summaries, doc edits, light refactors).
Constraints
- **No MCP tools** are available in background subagents.
- Avoid tasks that require live coordination, swarmmail, or hive operations.
- If MCP tools are needed, request a foreground worker instead.
Safe Use Cases
- Documentation updates
- Static file edits
- Copy edits and formatting
- Notes, summaries, and small refactors
Usage Guidance
If a task needs tool coordination or swarmmail calls, switch to a foreground worker.
Native Claude Code 2.1.32 Integration
Model
`model: haiku` - Background workers use the cheapest model since they handle simple, non-coordination tasks (doc edits, formatting, summaries).
Permission Mode
`permissionMode: acceptEdits` - Like foreground workers, background workers need to edit files autonomously.
Memory
`memory: local` - Background workers use local memory (session-scoped) instead of project memory since they don't participate in coordination or learning signals.
Hooks
Background workers can use hivemind CLI via Bash for learnings, but skip all MCP tool usage. The hooks provide graceful degradation when MCP is unavailable.
When to Use Background Workers
- Documentation updates that don't require tests or type checking
- Static file edits (JSON, YAML, config files)
- Copy edits, formatting, and style fixes
- Summaries and notes generation
**Never use background workers for:**
- Tasks requiring file reservations or swarmmail coordination
- Tasks requiring hive/swarm/hivemind MCP tools
- Complex refactors or multi-file changes with dependencies
Multi-agent coordination that survives context death. Break big tasks into small ones. Spawn agents to work in parallel. Learn from what works. swarmtools.ai · Documentation
Repo: joelhooks/swarm-tools

