New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

Launching Agent Teams

by lexler in lexler/skill-factory

Plans and launches Claude Code agent teams with distinct roles, right-sized tasks and detailed spawn prompts, and says when subagents or worktrees fit better.

Apache-2.0Auto-check passedAgent Workflows

Install Launching Agent Teams

skills CLI
$ npx skills add lexler/skill-factory --skill launching-agent-teams -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install lexler/skill-factory launching-agent-teams --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/lexler/skill-factory.git skills-src && mkdir -p .claude/skills && cp -r skills-src/output_skills/ai/claude-code/launching-agent-teams .claude/skills/launching-agent-teams && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
launching-agent-teams
GitHub stars
239
Token cost
~1.3k tokens
SKILL.md length
629 words
Files
4 (incl. scripts, references)
Skills in repo
25
Repo updated
First seen
Licence
Apache-2.0

At a glance

Plans and launches Claude Code agent teams with distinct roles, right-sized tasks and detailed spawn prompts, and says when subagents or worktrees fit better.

  • Works in 6 steps: Clarify the Work → Decompose Into Roles → Size Tasks Appropriately → …
  • Asking for a team of agents to work on separate parts of a project at once
  • SKILL.md covers Setup, Prerequisites, Teams vs Subagents — Pick the… and Launching a Team, plus 5 more sections
  • Runs Python scripts from its folder; calls python

What it does

Agent teams are experimental and need the CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS setting turned on, which the agent offers to add if it is missing. The skill first refreshes its reference docs with scripts/update-docs.py, then checks that a team is the right tool. Teams suit work where teammates must talk, debate or test competing hypotheses across independent areas, subagents suit focused tasks that only report back and cost fewer tokens, and git worktrees suit manual parallel sessions.

Launching a team means clarifying the goal, splitting it into non-overlapping roles with three to five teammates and around five or six tasks each, and sizing tasks as self-contained units with a clear deliverable. Teammates do not inherit the lead's conversation, so each spawn prompt names the files they own, the needed context, what done looks like and the conventions to follow. Warned-against patterns include two teammates editing one file and a teammate with a single tiny task. Reference notes on Anthropic's agent teams and subagents are included.

When your agent uses it

  • Asking for a team of agents to work on separate parts of a project at once
  • Deciding between an agent team, subagents and plain git worktrees
  • Writing spawn prompts that give each teammate the context it needs

Example prompts

  • “Create a team to migrate the frontend, backend and tests to the new auth API.”
  • “Should this refactor use an agent team or subagents?”
  • “Spawn teammates to investigate the flaky checkout test with competing hypotheses.”
  • “Set up three teammates for the docs overhaul and write their spawn prompts.”

Requirements

  • Claude Code with the experimental agent teams setting enabled
  • Python, to run the docs update script

Workflow steps

6 steps, taken from the step headings in SKILL.md.

  1. Clarify the Work
  2. Decompose Into Roles
  3. Size Tasks Appropriately
  4. Write Rich Spawn Prompts
  5. Configure Display Mode
  6. Create the Team

What it can do on your machine

Read from SKILL.md and the folder at commit e262032. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Ships 1 file in scripts/ (Python), which the agent can run.

    Shell commands in SKILL.md call:

    • python

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Launching Agent Teams loads about 1.3k tokens when it runs, and up to ~18k if it reads all its reference files. Until then it costs about 45 tokens; SKILL.md has 629 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~45
When it runs · the whole SKILL.md, loaded when a task matches
~1.3k
With references · SKILL.md plus every file in references/, read only if the agent opens them
~18k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); the scripts in this folder are not scanned.

SKILL.md

The full file from lexler/skill-factory at commit e262032, republished under its Apache-2.0 licence (© lexler). 629 words, ~1,311 tokens.

Download SKILL.mdSave it as .claude/skills/launching-agent-teams/SKILL.md (or your agent's skills folder). This skill also uses 3 other files; get the full folder from GitHub.
name
launching-agent-teams
description
Launches agent teams with structured roles and task decomposition. Use when asked to create a team, spawn teammates, or coordinate multiple agents in parallel.

STARTER_CHARACTER = 🚀👥

Setup

Update reference docs to get the latest from Anthropic:

bash
python ${CLAUDE_SKILL_DIR}/scripts/update-docs.py

Prerequisites

Agent teams are experimental. The user must have this in their settings (~/.claude/settings.json or .claude/settings.json):

json
{
  "env": {
    "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1"
  }
}

If the flag isn't set, tell the user and offer to add it before proceeding.

Teams vs Subagents — Pick the Right Tool

Before launching a team, confirm it's the right approach:

Use agent teams when:

  • Teammates need to communicate with each other
  • Work benefits from debate, challenge, or cross-pollination
  • Tasks span multiple independent areas (frontend/backend/tests)
  • Investigation needs competing hypotheses

Use subagents instead when:

  • Workers just report results back (no inter-agent communication)
  • Tasks are focused and self-contained
  • You want lower token cost
  • Sequential dependency between steps

Use git worktrees when:

  • You want manual parallel sessions without automated coordination

If the task fits subagents better, say so and offer that instead.

Launching a Team

1. Clarify the Work

Before spawning anything, understand:

  • What's the goal?
  • What are the independent pieces?
  • What would each teammate own?
2. Decompose Into Roles

Each teammate should own a distinct, non-overlapping area. Overlap causes file conflicts and wasted tokens.

Aim for 3-5 teammates. More than that creates coordination overhead that rarely pays off. Each teammate should have 5-6 tasks to stay productive.

Anti-patterns:

  • Two teammates editing the same file
  • A teammate with only one small task (just do it yourself)
  • Roles so broad they inevitably overlap
3. Size Tasks Appropriately
  • Too small: coordination overhead exceeds the benefit
  • Too large: teammates work too long without check-ins
  • Right size: self-contained units with a clear deliverable (a function, a test file, a review document)
4. Write Rich Spawn Prompts

Teammates do NOT inherit the lead's conversation history. They only get:

  • CLAUDE.md and project context
  • The spawn prompt you write

Include in each spawn prompt:

  • Specific files/modules they own
  • Relevant context they need (architecture decisions, constraints)
  • What "done" looks like
  • Technologies, patterns, or conventions to follow
5. Configure Display Mode
  • in-process (default): all teammates in one terminal, cycle with Shift+Down
  • tmux/auto: split panes, each teammate visible (requires tmux or iTerm2)

For monitoring multiple teammates, split panes are better. Suggest tmux if available.

6. Create the Team

Present the proposed team structure to the user for approval before spawning. The structure should cover:

  • Team goal
  • Number of teammates and their roles
  • What each teammate owns (specific files/modules)
  • Model choice (default: inherit from lead, or Sonnet for cost efficiency)
  • Whether plan approval is needed (recommend for risky/destructive work)
Show full SKILL.md (227 more words)Show less

Team Management

  • Plan approval: For risky work, require teammates to plan before implementing. The lead reviews and approves/rejects plans.
  • Direct messaging: Users can message individual teammates via Shift+Down (in-process) or clicking panes (split mode).
  • Task dependencies: Tasks can depend on other tasks. Blocked tasks auto-unblock when dependencies complete.
  • Hooks: TeammateIdle (exit code 2 sends feedback, keeps teammate working) and TaskCompleted (exit code 2 prevents completion).

Key Constraints

  • One team per session — clean up before starting a new one
  • No nested teams — teammates cannot spawn their own teams
  • Lead is fixed — can't transfer leadership
  • No session resumption for in-process teammates — after resume, spawn new ones
  • All teammates start with the lead's permission mode
  • Shutdown can be slow — teammates finish current tool call first

Parallel Work Principles

From large-scale multi-agent projects:

  • Decompose for true parallelism: if all agents converge on the same bug/file, parallelization fails regardless of team size
  • Design for agent orientation: teammates entering fresh need rich context in their spawn prompts and CLAUDE.md — documentation is their primary interface
  • Lightweight coordination: file-based task lists + git conflict resolution work better than complex orchestration
  • Acknowledge hard limits: autonomous agents have ceilings — plan for human verification of critical outcomes

Cleanup

Always clean up through the lead:

Clean up the team

Shut down teammates first, then clean up. If orphaned tmux sessions persist:

bash
tmux ls
tmux kill-session -t <session-name>

Reference

© lexler, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 3 other files (scripts, references) in output_skills/ai/claude-code/launching-agent-teams of lexler/skill-factory.

  • SKILL.md
  • references/anthropic-agent-teams.md
  • references/anthropic-sub-agents.md
  • scripts/update-docs.py

Open the folder on GitHubat commit e262032

Compare with similar skills

Launching Agent Teams next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Launching Agent Teams compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Launching Agent Teams this skilllexler/skill-factory239—~1.3kAutomated safety check: PassApache-2.0
OMA Multi-Agent Orchestratorfirst-fluke/oh-my-agent1.3k—~3.1kAutomated safety check: PassMIT
Swarm Parallel Dispatchlangchain-ai/langchain-skills1.3k—~3kAutomated safety check: PassMIT
Cursor Orchestratecursor/plugins11k—~1.1kAutomated safety check: PassNone
Agents Project Coordinatorasgeirtj/system_prompts_leaks69k—~2.9kAutomated safety check: PassCC0-1.0
Team Agent Pipelinezereight/gitlab-mcp2k1 repos~1.5kAutomated safety check: PassMIT

Similar skills

  • OMA Multi-Agent Orchestrator

    first-fluke/oh-my-agent

    Splits a complex feature into prioritized tasks, spawns specialist CLI subagents in parallel, tracks them through shared memory and verifies each result.

    1.3k GitHub stars~3.1k tokensUpdated today
    Agent WorkflowsAuto-check passed

  • Swarm Parallel Dispatch

    langchain-ai/langchain-skills

    Official

    Fans a list of independent items out to subagents in parallel, merges the results back into a table and supports retrying only the rows that failed.

    1.3k GitHub stars~3k tokensUpdated yesterday
    Agent WorkflowsAuto-check passed

  • Cursor Orchestrate

    cursor/plugins

    Official

    Splits a large goal into a tree of parallel Cursor cloud agents, with planners, workers and verifiers coordinated by a script and reporting through structured handoffs.

    11k GitHub stars~1.1k tokensUpdated today
    Agent WorkflowsAuto-check passed

  • Agents Project Coordinator

    asgeirtj/system_prompts_leaks

    Runs a goal as a project in which the agent coordinates separate agent threads, judging when to split the work, and interviews you first when nothing can be verified.

    69k GitHub stars~2.9k tokensUpdated today
    Agent WorkflowsAuto-check passed

  • Team Agent Pipeline

    zereight/gitlab-mcp

    Spawns a chosen number of coordinated agents on a shared task list and runs them through plan, requirements, execution, verification and fix stages.

    2k GitHub starsUsed in 1 repo~1.5k tokens
    Agent WorkflowsAuto-check passed

  • Agent Orchestrator

    NeverSight/learn-skills.dev

    Meta-agent skill for orchestrating complex tasks through autonomous sub-agents.

    217 GitHub starsUsed in 1 repo~1.4k tokens
    Agent WorkflowsAuto-check passed

More from lexler/skill-factory

All 25 skills in this repo

  • C4 Architecture Diagrams

    lexler/skill-factory

    Creates C4 model diagrams at every zoom level, from system landscape to code, in ASCII, Mermaid or Structurizr, for designing or documenting software architecture.

    239 GitHub stars~2.3k tokensUpdated today
    Auto-check passed

  • Claude Code Statusline Writer

    lexler/skill-factory

    Guides writing and debugging Claude Code status line scripts that read session JSON from stdin and print one line of text.

    239 GitHub stars~872 tokensUpdated today
    Auto-check passed

  • Augmented Coding Patterns Reference

    lexler/skill-factory

    Catalog of obstacles, anti-patterns and patterns for working with AI coding agents, covering context management and reliability, from a published patterns collection.

    239 GitHub stars~1.5k tokensUpdated today
    Auto-check passed

  • Approval Testing Toolkit

    lexler/skill-factory

    Writes snapshot-style approval tests in Python, JavaScript, TypeScript or Java, comparing output against an approved file instead of writing individual assertions.

    239 GitHub stars~1.2k tokensUpdated today
    Auto-check passed

  • Hotspots

    lexler/skill-factory

    Find where a codebase actually costs time by mining its git history (Tornhill hotspot analysis).

    239 GitHub stars~1.2k tokensUpdated today
    Auto-check passed

  • Nullables Testing Pattern

    lexler/skill-factory

    Teaches the Nullables pattern for testing without mocking libraries: production classes with an off switch, stubbed only at the third-party edge.

    239 GitHub stars~2.2k tokensUpdated today
    Auto-check passed

Categories

Questions about Launching Agent Teams

What does Launching Agent Teams do?

Plans and launches Claude Code agent teams with distinct roles, right-sized tasks and detailed spawn prompts, and says when subagents or worktrees fit better. Agent teams are experimental and need the CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS setting turned on, which the agent offers to add if it is missing.py, then checks that a team is the right tool.

When should I use Launching Agent Teams?

Launching Agent Teams fits situations like: asking for a team of agents to work on separate parts of a project at once; deciding between an agent team, subagents and plain git worktrees; writing spawn prompts that give each teammate the context it needs.

How do I install Launching Agent Teams in Claude Code?

Run `npx skills add lexler/skill-factory --skill launching-agent-teams -a claude-code`. Or copy the skill folder (output_skills/ai/claude-code/launching-agent-teams in lexler/skill-factory) into .claude/skills/launching-agent-teams in your project. Claude Code loads it when a task matches its description.

How do I install Launching Agent Teams in Codex?

Run `npx skills add lexler/skill-factory --skill launching-agent-teams -a codex`. Or copy the skill folder (output_skills/ai/claude-code/launching-agent-teams in lexler/skill-factory) into .agents/skills/launching-agent-teams in your project. Codex loads it when a task matches its description.

Can I use Launching Agent Teams in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add lexler/skill-factory --skill launching-agent-teams -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/launching-agent-teams, .gemini/skills/launching-agent-teams, .github/skills/launching-agent-teams and .opencode/skills/launching-agent-teams in your project.

What does Launching Agent Teams need to run?

Going by SKILL.md and its folder, Launching Agent Teams needs Python for the scripts in its folder and the command-line tools its instructions call (python). Our summary lists: Claude Code with the experimental agent teams setting enabled; Python, to run the docs update script.

Does Launching Agent Teams access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Launching Agent Teams safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. The check reads SKILL.md only: the scripts in the folder are not scanned, so read them before running anything.

What licence does Launching Agent Teams use?

Launching Agent Teams is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Launching Agent Teams use?

About 1.3k tokens (SKILL.md is roughly 5.2k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full. Its references folder adds about 17k tokens, read only when the agent opens those files.

What are the alternatives to Launching Agent Teams?

Skills that share tags, products or a category with Launching Agent Teams: OMA Multi-Agent Orchestrator (first-fluke/oh-my-agent, 1.3k stars), Swarm Parallel Dispatch (langchain-ai/langchain-skills, 1.3k stars), Cursor Orchestrate (cursor/plugins, 11k stars) and Agents Project Coordinator (asgeirtj/system_prompts_leaks, 69k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Launching Agent Teams?

lexler (a GitHub user) maintains it in lexler/skill-factory, which has 239 GitHub stars. The repository holds 25 skills in this directory. The repository was last updated on October 10, 2026.

Source: lexler/skill-factory on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.