team

Coordinates multi-agent sub-agent teams to implement multi-task specs through the SDLC.

Updated Nov 2, 2025
One-click install
npx skills add https://github.com/fx/skills --skill team-fx
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: team
Source: https://github.com/fx/skills/tree/main/skills/team
Command: npx skills add https://github.com/fx/skills --skill team-fx

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Implementing a large spec or multi-task feature with a single agent leads to context exhaustion, skipped lifecycle stages, and conflicting edits. This Skill turns the main session into a strict coordinator that spawns focused sub-agents, isolates their work, and drives every task through review, CI, and merge. ## Core Features & Use Cases - Coordinator-driven SDLC: The main session delegates only single-purpose jobs (coding, verification, fixes) while owning review triage, CI monitoring, and merge gates itself. - Worktree isolation: Pre-creates one git worktree per concurrent coder with a mandatory prompt preamble, preventing parallel agents from corrupting a shared working tree. - Scope Brief discipline: Captures the user's verbatim request and out-of-scope items, then injects it into every coder and reviewer prompt so findings are judged against what was actually asked. - Use Case: Given a change document with six tasks, the coordinator plans waves, spawns worktree-pinned coder agents in parallel, runs reviewer and CI waiters, resolves feedback, and merges each PR through a mandatory gate checklist. ## Quick Start Ask the agent to use the team skill to implement all tasks in your spec file as a coordinated multi-agent workflow.

Frequently Asked Questions about team

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run a multi-agent implementation workflow on a spec?▼

Invoke the team skill with a spec file, issue URL, or task description. The coordinator breaks the work into tasks, spawns focused coder agents in isolated git worktrees, and drives each resulting PR through review, CI, and merge gates.

How do I run parallel coding agents without them conflicting?▼

Pre-create one git worktree per concurrent coder on its own branch, then pin each agent to its worktree with a prompt preamble forcing absolute paths. The isolation flag on the spawn call is a no-op for teammates, so real worktrees are mandatory.

Does the team skill work on both Claude Code and Codex?▼

Yes, but the mechanics differ per host. Claude Code teams form implicitly on the first agent spawn and notify on completion, while Codex requires fork_turns set to none, explicit wait_agent calls before the turn ends, and lowercase task_name handles.

Why did my spawned teammate stop responding mid-task?▼

A teammate showing idle status has ended its turn and will not resume until messaged; only a completed status settles a wait. On Codex, ending the coordinator turn also terminates all live teammates, so you must wait on them first.

When should I not use a coordinated agent team?▼

Avoid it for single-task or small changes where one coder agent suffices, and never let the coordinator write code itself. The skill is explicit-use only and activates when the user names it or an active workflow calls it.