/agent-team
Orchestrating Claude Code agent teams. Use when creating teams, spawning teammates, assigning tasks, configuring teammate modes, or setting up team quality gate hooks.
$ npx -y skills add bendrucker/claude --skill agent-team --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
- Fires itselfAuto-invocation. Claude auto-loads it when your prompt matches the work.
- You can call itInvoke it directly when you want it.
- Slash command
/agent-team
Context preview
The summary Claude sees to decide when to auto-load this skill.
Orchestrating Claude Code agent teams. Use when creating teams, spawning teammates, assigning tasks, configuring teammate modes, or setting up team quality gate hooks.
SKILL.md
agent-team.SKILL.mdname: claude-code:agent-team
description: Orchestrating Claude Code agent teams. Use when creating teams, spawning teammates, assigning tasks, configuring teammate modes, or setting up team quality gate hooks.
argument-hint: "[--plan-approval] [--delegate] [--shutdown]"
allowed-tools:
- Read
- Write
- Edit
- Glob
- Grep
- Bash
- WebFetch(domain:code.claude.com)
Agent Teams
Arguments
These select a team mode up front instead of toggling it interactively (see [Team Lifecycle](#team-lifecycle)):
- `--delegate`: start the lead in delegate mode, coordination-only with no direct implementation. Default: the lead can implement.
- `--plan-approval`: require teammates to plan before implementing, read-only until the lead approves. Default: off.
- `--shutdown`: shut the current team down, teammates first and then the team, then stop. Default: create or operate a team.
Team Lifecycle
Describe the task and team structure in natural language. Claude creates the team, spawns teammates, and coordinates work. Delegate mode also toggles interactively with Shift+Tab. Always shut down via the lead: teammates individually, then the team.
Task Management
Tasks: pending, in progress, completed. Tasks can depend on other tasks — blocked tasks unlock automatically when dependencies complete.
- **Lead assigns**: tell the lead which task to assign
- **Self-claim**: teammates pick up unassigned, unblocked tasks after finishing
- File locking prevents race conditions on simultaneous claims
Communication
- **message**: send to one specific teammate
- **broadcast**: send to all teammates (use sparingly — costs scale with team size)
- Messages deliver automatically; the lead doesn't need to poll
- Idle notifications are automatic when a teammate finishes a turn
Architecture
Teams are stored locally:
- **Team config**: `~/.claude/teams/{team-name}/config.json`
- **Task list**: `~/.claude/tasks/{team-name}/`
The config contains a `members` array with each teammate's name, agent ID, and type. Teammates read this file to discover each other.
Teammates inherit the lead's permission settings at spawn. Each loads project context (CLAUDE.md, MCP servers, skills) plus the spawn prompt — the lead's conversation history does not carry over.
Limitations
- No session resumption for in-process teammates (`/resume` won't restore them)
- Task status can lag — check and nudge if stuck
- One team per session, no nested teams
- Lead is fixed for the team's lifetime
Known Issues
- **tmux backend**: teammates spawned in tmux panes may never receive their initial prompt or poll their inbox, causing them to idle indefinitely ([#23415](https://github.com/anthropics/claude-code/issues/23415)). A separate race condition between pane creation and shell initialization can prevent the `claude` command from executing at all ([#25315](https://github.com/anthropics/claude-code/issues/25315)). Use `in-process` backend (the default) as a workaround.
References
- [references/hooks.md](references/hooks.md): `TeammateIdle` and `TaskCompleted` quality gate hooks
- [references/practices.md](references/practices.md): task sizing, when to use teams, decomposition, model selection, worktree dispatch
- [references/interaction.md](references/interaction.md): display modes, keyboard shortcuts, direct teammate interaction
- [Agent Teams Documentation](https://code.claude.com/docs/en/agent-teams)
Read more
name: claude-code:agent-team description: Orchestrating Claude Code agent teams. Use when creating teams, spawning teammates, assigning tasks, configuring teammate modes, or setting up team quality gate hooks. argument-hint: "[--plan-approval] [--delegate] [--shutdown]" allowed-tools: - Read - Write - Edit - Glob - Grep - Bash - WebFetch(domain:code.claude.com)
Agent Teams
Arguments
These select a team mode up front instead of toggling it interactively (see [Team Lifecycle](#team-lifecycle)):
- `--delegate`: start the lead in delegate mode, coordination-only with no direct implementation. Default: the lead can implement.
- `--plan-approval`: require teammates to plan before implementing, read-only until the lead approves. Default: off.
- `--shutdown`: shut the current team down, teammates first and then the team, then stop. Default: create or operate a team.
Team Lifecycle
Describe the task and team structure in natural language. Claude creates the team, spawns teammates, and coordinates work. Delegate mode also toggles interactively with Shift+Tab. Always shut down via the lead: teammates individually, then the team.
Task Management
Tasks: pending, in progress, completed. Tasks can depend on other tasks — blocked tasks unlock automatically when dependencies complete.
- **Lead assigns**: tell the lead which task to assign
- **Self-claim**: teammates pick up unassigned, unblocked tasks after finishing
- File locking prevents race conditions on simultaneous claims
Communication
- **message**: send to one specific teammate
- **broadcast**: send to all teammates (use sparingly — costs scale with team size)
- Messages deliver automatically; the lead doesn't need to poll
- Idle notifications are automatic when a teammate finishes a turn
Architecture
Teams are stored locally:
- **Team config**: `~/.claude/teams/{team-name}/config.json`
- **Task list**: `~/.claude/tasks/{team-name}/`
The config contains a `members` array with each teammate's name, agent ID, and type. Teammates read this file to discover each other.
Teammates inherit the lead's permission settings at spawn. Each loads project context (CLAUDE.md, MCP servers, skills) plus the spawn prompt — the lead's conversation history does not carry over.
Limitations
- No session resumption for in-process teammates (`/resume` won't restore them)
- Task status can lag — check and nudge if stuck
- One team per session, no nested teams
- Lead is fixed for the team's lifetime
Known Issues
- **tmux backend**: teammates spawned in tmux panes may never receive their initial prompt or poll their inbox, causing them to idle indefinitely ([#23415](https://github.com/anthropics/claude-code/issues/23415)). A separate race condition between pane creation and shell initialization can prevent the `claude` command from executing at all ([#25315](https://github.com/anthropics/claude-code/issues/25315)). Use `in-process` backend (the default) as a workaround.
References
- [references/hooks.md](references/hooks.md): `TeammateIdle` and `TaskCompleted` quality gate hooks
- [references/practices.md](references/practices.md): task sizing, when to use teams, decomposition, model selection, worktree dispatch
- [references/interaction.md](references/interaction.md): display modes, keyboard shortcuts, direct teammate interaction
- [Agent Teams Documentation](https://code.claude.com/docs/en/agent-teams)
My personal plugin marketplace for Claude Code, Anthropic's AI coding assistant.
Repo: bendrucker/claude
Other skills on bendrucker-claude.
- /agent-ideas
Harvest agent-tooling ideas from prominent developers.
Open skill - /cleye
Type-safe CLI argument parsing with cleye, the standard parser for this repo's Bun scripts. Use when writing or editing any script that takes arguments (flags, positional parameters, subcommands, --help) instead of reading existing scripts for the pattern.
Open skill - /coverage
Measure Bun test coverage and close gaps on a specific file. Use when adding or editing tests, when asked about coverage, or when the PostToolUse coverage hook reports uncovered lines.
Open skill - /activity
Report real device usage from ActivityWatch. Covers per-app time, window titles, and active vs idle spans. Use when asked "what apps did I use", "how long was I in X", "what did I work on today", "how much was I active vs idle", or to mine usage patterns for automation.
Open skill - /history
Report shell history from atuin's local capture. Covers what commands ran, when, where, and how they exited. Use when asked "what commands did I run", "what was I working on in the terminal", "have I ever run X", "how do I usually invoke X", or about recent shell activity,
Open skill - /bun
Bun runtime patterns. Use when running bun commands, working with package.json/bun.lock, writing TypeScript scripts under Bun, or developing Claude Code plugins.
Open skill

