PlugWiki

team-composition-patterns Skill (wshobson) for Claude: What It Does and How to Install It

Helps Claude Code decide how many agents to spawn for a job and which kind each one should be, for developers using Claude Code's Agent Teams.

Installs
9.8K
on skills.sh
Stars
42.1K
on GitHub
Files
2
besides SKILL.md
Updated
Oct 2026
last repo commit

team-composition-patterns gives Claude a set of sizing rules and seven preset teams for Claude Code's Agent Teams feature. It covers the review, debug, feature, fullstack, research, security and migration teams. It also says which subagent type each role needs, because read-only types such as Explore and Plan cannot edit files.

It loads when you ask how many agents to spawn, which team to use, which display mode to pick (tmux, iTerm2 or in-process), or how to build a custom team. The skill is guidance only. Its team types come from the agent-teams plugin in the same repo, so install the plugin to get them. The plugin README also asks you to set an experimental environment variable first.

Good for

  • Claude Code users running several agents at once on one task
  • Developers choosing between a review, debug, feature or migration team
  • Teams running agents in CI who need the in-process display mode

Install team-composition-patterns

Claude Code

  1. 1

    Copy this prompt.

    Install the team-composition-patterns skill in Claude Code. Run this command:

/plugin marketplace add wshobson/agents
/plugin install agent-teams@claude-code-workflows

Then check that it loaded and tell me how to use it.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

Requires Claude Code.

The plugin route also gives you the agent-teams:team-lead, team-reviewer, team-debugger and team-implementer types this skill tells Claude to use, plus 7 commands such as /team-spawn. Set CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 first, as the plugin README asks. The skills CLI route installs this skill alone, without those agent types.

From the project README. Source.

What's in the skill

When it loads

Design optimal agent team compositions with sizing heuristics, preset configurations, and agent type selection. Use this skill when deciding how many agents to spawn for a task, when choosing between a review team versus a feature team versus a debug team, when selecting the correct subagent_type for each role to ensure agents have the tools they need, when configuring display modes (tmux, iTerm2, in-process) for a CI or local environment, or when building a custom team composition for a non-standard workflow such as a migration or security audit.

Sections in SKILL.md

  1. When to Use This Skill
  2. Team Sizing Heuristics
  3. Preset Team Compositions
  4. Review Team
  5. Debug Team
  6. Feature Team
  7. Fullstack Team
  8. Research Team
  9. Security Team
  10. Migration Team
  11. Agent Type Selection
  12. Display Mode Configuration
  13. Custom Team Guidelines
  14. Troubleshooting
  15. Related Skills

Key facts

Author
Seth Hobson (wshobson), listed as author of the agent-teams plugin
Repo
wshobson/agents, plugin agent-teams
License
MIT (plugin.json)
Installs on skills.sh
9,821 (Nov 2, 2026)
GitHub stars
42,081 for the whole repo (Nov 2, 2026)
Last commit
Oct 30, 2026 (skill folder)
Files
3 files including SKILL.md: two reference guides and no scripts
Needs
The experimental Agent Teams feature, switched on with an environment variable
Plugin bundles
6 skills, 4 agents and 7 commands
skills.sh audits
Gen Agent Trust Hub: pass; Socket: pass; Snyk: pass (Nov 2, 2026)

What it includes

  • Maps task complexity to a team of 1 to 5 agents

    Four levels from simple (1-2 agents) to very complex (4-5). The rule of thumb is the smallest team that covers every dimension.

    Team sizing heuristics

  • Seven ready-made teams: review, debug, feature, fullstack, research, security, migration

    Each preset lists its size, its agent types and when to use it. The review team is 3 reviewers; the migration team is 1 lead, 2 implementers and 1 reviewer.

    Preset teams

  • Says which subagent_type gives a teammate the tools it needs

    Explore and Plan are read-only, so they never get implementation tasks. general-purpose and the agent-teams types can change files or run commands.

    Agent type selection

  • Three general-purpose agents that search code and the web in parallel

    The skill lists Grep, Glob and Read plus WebSearch and WebFetch as this team's capabilities.

    Network · Research team

  • Shows the teammateMode setting for tmux, iterm2 or in-process

    You put the setting in ~/.claude/settings.json yourself. In-process is the one suggested for CI.

    Display mode setup

  • Five rules: coordinator, matched roles, no duplicates, clear boundaries, small size

    Two to four teammates is called the sweet spot; five or more needs a lot of coordination.

    Custom team guidelines

  • Five fixes, from read-only agents to teammates that receive no tasks

    For example, change an Explore teammate to general-purpose when it has to write files.

    Troubleshooting

  • Longer guide to choosing agent types

    Bundled reference file the skill points to.

    Read · references/agent-type-selection.md

  • Full preset definitions with member names and task templates

    Bundled reference file the skill points to.

    Read · references/preset-teams.md

From the official docs

Permissions

Medium risk

The skill is advice, but the agents it picks can edit files and run shell commands. Check each teammate's file ownership before you let a team start.

What it can do

  • Read your data
  • Create and change things
  • Delete things
  • Read-only mode

Watch out for

  • The agents it picks can edit files and run commands

    The agent-teams plugin lists Write, Edit and Bash for team-implementer, and Bash for the other three agents. Review the file ownership each teammate is given before you approve edits. github.com

  • It needs an experimental feature switched on

    The plugin README says to set CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 and to put a teammateMode setting in ~/.claude/settings.json. You make both changes yourself. github.com

  • The plugin route adds more than this skill

    The plugin installs 6 skills, 4 agents and 7 commands. It has no hooks and no bundled MCP server in the git tree. The skills CLI route installs only this skill. github.com

  • More agents means more usage

    The skill itself says that adding teammates increases coordination overhead. Start with the smallest team that covers the job. github.com

Technical details
Turning on read-onlyNo read-only mode. The Explore and Plan agent types are read-only, and the skill recommends them for research and planning.
Sign-in and keysSKILL.md names no key or token.
Scopes

  • Spawns Claude Code agents; implementer agents can edit files and run shell commands in your project

Questions

What does team-composition-patterns do?

It tells Claude how many agents to spawn for a task and which type each one should be. It includes seven preset teams, a sizing table and display mode settings for Claude Code's Agent Teams feature.

Is this skill made by Anthropic?

No. It is a community skill in the wshobson/agents repo, listed under Seth Hobson in the plugin manifest. The Agent Teams feature it describes belongs to Claude Code.

How do I install team-composition-patterns?

Use the install box above. The plugin route adds this skill and the agent types it refers to. The skills CLI route adds only this skill.

Do I need the plugin, or is the skill enough?

The skill names subagent types such as agent-teams:team-reviewer. Those come from the plugin, so install the plugin if you want to use them. The skill alone still gives the sizing rules and the presets as advice.

What do I have to switch on first?

The plugin README says to set the environment variable CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS to 1. It also suggests a teammateMode value in ~/.claude/settings.json, with tmux, iterm2 or in-process as the options.

Does it work in the Claude app or Codex?

There is no Claude app upload tab, because the skill is about Claude Code's Agent Teams and Agent tool. The repo documents a Codex plugin route and the plugin has a Codex manifest, but the skill text is written for Claude Code.

Is team-composition-patterns safe?

The skill is a text guide with two reference files and no scripts. The agents it recommends can run shell commands, and team-implementer can edit files, so check what each teammate owns before you start a team.

How is it different from parallel-feature-development?

team-composition-patterns picks the team size and agent types. Its SKILL.md points to parallel-feature-development for splitting work streams and file ownership once the team exists, and to team-communication-protocols for messaging and shutdown.

Information

updated Nov 2, 2026

Repository
wshobson/agents
Folder
plugins/agent-teams/skills/team-composition-patterns

wshobson/agentsskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 10
  1. agent-teams plugin README · github.com
  2. wshobson/agents README · github.com
  3. agent-teams Codex plugin.json · github.com
  4. SKILL.md · github.com
  5. agent-teams plugin.json · github.com
  6. skills.sh listing · skills.sh
  7. agent-type-selection.md · github.com
  8. preset-teams.md · github.com
  9. team-implementer.md · github.com
  10. team-lead.md · github.com

Updated Nov 2, 2026.