cubetribe/claudecode_godmode-on/agent-teams

Claude Code and agent skill by cubetribe/claudecode_godmode-on

Trust tier 2
  • community
LLM review pending
Source check pending

Checking that the SKILL.md file is still in the repository.

Experimental Agent Teams orchestration — run CC_GodMode agents as parallel teammates with SharedTaskList coordination (requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1)

Install

npx vskill@latest install cubetribe/claudecode_godmode-on/agent-teams

SKILL.md

Agent Teams Orchestration (Experimental)

Status: Experimental — OFF by default (requires CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1)
When to use: Large features where multiple teammates can work truly in parallel and the user explicitly accepts the token cost

What Are Agent Teams?

Agent Teams are coordinated background sessions with a shared task list, inter-agent messaging, and a lead/manager that assigns tasks and keeps workers in sync. Instead of the Orchestrator calling agents sequentially via Task tool, a lead agent dispatches work and teammates claim, execute, and report tasks from the shared queue independently.

This surface is experimental and off by default — enable it only by setting CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 in your environment or settings.json.

How This Relates to the Other Parallelization Surfaces

GodMode maps onto four parallelization surfaces. Choose the right one for the job:

SurfaceWhat it isUse when
Plain subagents (Task tool)Delegated workers inside one session, own context, return a summaryA side task would flood the main context; up to ~10 run concurrently, the rest queue
Agent teams (this skill)Coordinated sessions, shared task list, inter-agent messaging, lead-managedSplit a project, assign pieces, keep workers in sync — persistent teammates, not one-off tasks
Dynamic workflows (skills/dynamic-workflows/)Orchestrator-written script fans out to tens–hundreds of subagents + adversarial cross-checksJob outgrows a handful of subagents: codebase-wide audit, large migration, cross-checked research
Agent view (claude agents)Dispatch + monitor background sessionsSeveral independent hand-off tasks you check on later (research preview)

Decision guidance:

  • Use plain subagents for one-off in-session side tasks (the default, always available).
  • Use agent teams when you need persistent, coordinated teammates kept in sync via a shared task list across a multi-part project.
  • Use dynamic workflows when a job outgrows a handful of subagents and needs fan-out at scale plus adversarial verification — trigger with the word "workflow" in a prompt or switch on ultracode.
  • See also the ## Parallelization section of CLAUDE.md for the cost guardrail and concurrency tiers that apply across all surfaces.

Enabling Agent Teams

Add to your environment or settings.json:

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

Team Composition

TeammateRoleModelIsolation
ArchitectDesign & planningopus—
BuilderImplementationsonnetworktree
TesterUX quality (when ux_gate: auto)sonnetworktree
SecuritySecurity gate (on security surfaces)sonnetworktree
ScribeDocumentationhaiku—

Recommended team size: 3–5 teammates for optimal coordination.

SharedTaskList

Tasks follow this lifecycle:

pending → in_progress → completed
→ failed
  • Dependencies: Tasks can have blockedBy arrays
  • Claiming: File-locking prevents race conditions
  • Auto-pickup: After completing a task, teammate picks next unblocked task

Example Task Flow

json
{
"tasks": [
{"id": "1", "name": "Design auth module", "assignee": "architect", "status": "completed"},
{"id": "2", "name": "Implement auth API", "assignee": "builder", "blockedBy": ["1"], "status": "in_progress"},
{"id": "3", "name": "Implement auth UI", "assignee": "builder", "blockedBy": ["1"], "status": "in_progress"},
{"id": "4", "name": "Security-check auth code", "assignee": "security", "blockedBy": ["2", "3"], "status": "pending"},
{"id": "5", "name": "Test auth UX", "assignee": "tester", "blockedBy": ["2", "3"], "status": "pending"}
]
}

Hooks for Agent Teams

HookPurposeExit Code 2
TeammateIdleFires when teammate about to go idleTeammate continues working
TaskCompletedFires when task marked completePrevents completion, sends feedback

When to Use Agent Teams vs Standard Orchestration

ScenarioUse Agent Teams?
Large feature with independent modulesYES
Multiple parallel investigationsYES
Small bug fixNO — overhead not worth it
API change (strict sequence required)NO — sequential workflow needed
Research taskNO — single agent sufficient

Limitations

  • No session resumption with in-process teammates
  • Task status can lag
  • Only one team per session
  • No nested teams
  • Token usage: ~15x standard
  • Shutdown can be slow

Use skills/departments/ first when the problem is ownership and routing. Use Agent Teams only when the implementation work itself can safely run in parallel.

Best Practices

  1. Break work into truly independent tasks
  2. Define clear dependencies with blockedBy
  3. Keep team size to 3–5 teammates
  4. Use worktree isolation for all building/testing teammates
  5. Reserve standard orchestration for sequential workflows (API changes)

CC_GodMode — © 2025–2026 Dennis Westermann ([dennis-westermann.de](https://www.dennis-westermann.de)). Proprietary — not open source. Free for private, non-commercial use; redistribution or re-hosting outside GitHub is prohibited; attribution required. Official source: [github.com/cubetribe/ClaudeCode_GodMode-On](https://github.com/cubetribe/ClaudeCode_GodMode-On). See LICENSE.

Security

  • Security scanPassed, 100 of 100Pattern checks on the skill's files.
  • Model reviewNot runA model review is scheduled for the next rescan.

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

Repository stars
53
Repository forks
10
7-day trend score
18

Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/cubetribe/claudecode_godmode-on/agent-teams/badge)](https://verified-skill.com/skills/cubetribe/claudecode_godmode-on/agent-teams)

Something wrong with this skill or its page? Tell us.