team-composition-patterns_skill

This skill helps you design optimal multi-agent team configurations by sizing, presets, and agent-type selection for Claude Code workflows.

Python

28.7k

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill wshobson/agents --skill team-composition-patterns

SKILL.md5.9 KB

Overview

This skill designs optimal multi-agent team compositions for Claude Code workflows. It provides sizing heuristics, preset team configurations, and guidance for selecting agent types and display modes. Use it to quickly configure efficient, role-aligned agent teams for implementation, review, debugging, research, and migrations.

How this skill works

The skill inspects task complexity and recommended dimensions to suggest team sizes and presets (Review, Debug, Feature, Fullstack, Research, Security, Migration). It maps roles to agent types (general-purpose, Explore, Plan, team-reviewer, team-debugger, team-implementer, team-lead) and recommends display modes (tmux, iTerm2, in-process) for runtime setup. It also provides rules for custom teams: coordinator assignment, role boundaries, and avoiding duplicate responsibilities.

When to use it

Deciding how many teammates to spawn for a specific task
Choosing a preset composition for code review, debugging, feature work, or migration
Selecting subagent_type when spawning teammates with the Task tool
Configuring teammate display mode for local development or CI
Designing a custom team for cross-cutting or full-stack work

Best practices

Start with the smallest team that covers required dimensions; add members only as needed
Assign a coordinator (team-lead) or clear user coordination for non-trivial teams
Match roles to agent types and never assign implementation to read-only agents
Define file ownership and responsibilities upfront to avoid overlap
Keep teams between 2–4 members for most tasks; 5+ needs explicit orchestration

Example use cases

Run a 3-member Review Team (security, performance, architecture) for a large PR
Spawn a 3-agent Debug Team to test three competing hypotheses on a flaky failing test
Assemble a Fullstack Team with lead + frontend, backend, and test implementers for a cross-layer feature
Use a Research Team of Explore agents to parallelize codebase search and web research on libraries
Configure teammateMode to tmux for local development monitoring or in-process for CI lightweight runs

FAQ

Estimate task complexity: simple (1–2), moderate (2–3), complex (3–4), very complex (4–5). Start small and expand if coordination gaps appear.

Which agent type should do file edits?

Use general-purpose or team-implementer/team-lead; read-only types (Explore, Plan) cannot modify files.

When should I use tmux vs in-process display modes?

Use tmux for interactive development and monitoring multiple agents, iTerm2 for macOS preference, and in-process for CI or simple scripted runs.