team
N coordinated agents on shared task list with staged pipeline. Activate when user says: team, spawn agents, parallel agents, coordinate agents, multi-agent, swarm, work together.
How do I install this agent skill?
npx skills add https://github.com/zereight/gitlab-mcp --skill teamIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill provides a framework for coordinating multiple AI agents across a staged pipeline (planning, execution, and verification). It handles task decomposition and state management using local file storage. No malicious patterns or security risks were identified.
- Socketpass
No alerts
- Snykpass
Risk: LOW · No issues
What does this agent skill do?
Team
Spawn N coordinated agents working on a shared task list. Uses VS Code's native subagent system for team management, inter-agent messaging, and task dependencies.
When to Use
- Task is decomposable into independent subtasks
- Multiple files/modules need parallel work
- Large-scale refactoring, migration, or multi-service work
When NOT to Use
- Single-file changes → use
/omg-autopilotor direct editing - Sequential pipeline → use
/omg-autopilot - Just need a plan → use
/planor/ralplan
Usage
/team 3:executor "fix all TypeScript errors"
/team 4:designer "implement responsive layouts"
/team "refactor the auth module"
/team ralph "build a complete REST API"
Parameters
- N — Number of agents (1-20). Defaults to auto-sizing.
- agent-type — Agent for
team-execstage (executor, debugger, designer, etc.). Defaults to stage-aware routing. - ralph — Wraps team in Ralph's persistence loop (retry on failure, verification before completion).
- task — High-level task to decompose and distribute.
Architecture
User: "/team 3:executor fix all TypeScript errors"
|
v
[omg-coordinator (Lead)]
|
+-- Analyze & decompose → subtask list
|
+-- Create tasks with dependencies
|
+-- Spawn N worker agents (subagents)
|
+-- Monitor loop (messages + task polling)
|
+-- Completion → shutdown workers → cleanup
Staged Pipeline
team-plan → team-prd → team-exec → team-verify → team-fix (loop)
Stage Agent Routing
Each stage uses specialized agents — not just executors:
| Stage | Required Agents | Optional Agents |
|---|---|---|
| team-plan | @explore, @planner | @analyst, @architect |
| team-prd | @analyst | @critic |
| team-exec | @executor | @debugger, @designer, @writer, @test-engineer |
| team-verify | @verifier | @test-engineer, @security-reviewer, @code-reviewer |
| team-fix | @executor | @debugger |
Routing rules:
- Lead picks agents per stage, not the user. User's
N:agent-typeonly overridesteam-execworkers. - Specialist agents complement executors. Route analysis to @architect, UI to @designer.
- Security-sensitive or >20 file changes must include @security-reviewer + @code-reviewer in
team-verify.
Stage Entry/Exit Criteria
- team-plan: Entry = invocation. Exit = task graph ready.
- team-prd: Entry = scope ambiguous. Exit = acceptance criteria explicit.
- team-exec: Entry = tasks created + workers spawned. Exit = tasks reach terminal state.
- team-verify: Entry = execution pass done. Exit (pass) = gates pass. Exit (fail) = fix tasks generated.
- team-fix: Entry = defects found. Exit = fixes done, return to team-verify.
Verify/Fix Loop
Continue team-exec → team-verify → team-fix until:
- Verification passes with no required fixes, or
- Max fix attempts exceeded (default: 3) → terminal
failed
Workflow
Phase 1: Parse Input
- Extract N (agent count, validate 1-20)
- Extract agent-type
- Extract task description
- Check for
ralphmodifier
Phase 2: Analyze & Decompose
Use @explore or @architect to analyze the codebase and break the task into N subtasks:
- Each subtask should be file-scoped or module-scoped
- Subtasks must be independent or have clear dependencies
- Each needs a concise subject and detailed description
Phase 3: Create Team
Write team state via omg_write_state:
{
"mode": "team",
"active": true,
"team_name": "fix-ts-errors",
"agent_count": 3,
"current_phase": "team-plan",
"task": "fix all TypeScript errors"
}
Phase 4: Create Tasks
Create subtasks with dependencies. Pre-assign owners to avoid race conditions.
Phase 5: Spawn Workers
Spawn N subagents in parallel using the agents field references:
- Each worker gets team preamble + assigned tasks
- Workers execute independently and report back
Phase 6: Monitor
Monitor via two channels:
- Inbound messages — workers report completion or need help
- Task polling — check overall progress periodically
Coordination actions:
- Unblock a worker with guidance
- Reassign work if a worker finishes early
- Handle failures — reassign or spawn replacement
Watchdog policy:
- Task stuck >5 min without messages → send status check
- No messages >10 min → reassign task
- Worker fails 2+ tasks → stop assigning to it
Phase 7: Stage Transitions
Update state on every stage transition via omg_write_state.
Phase 8: Completion
- Signal shutdown to all workers
- Wait for responses (15s timeout)
- Clean up team resources and state
- Report summary
Stage Handoff Convention
Each stage produces a handoff document before transitioning:
## Handoff: <current-stage> → <next-stage>
- **Decided**: [key decisions]
- **Rejected**: [alternatives and why]
- **Risks**: [identified risks]
- **Files**: [key files modified]
- **Remaining**: [items for next stage]
Handoffs saved to .omc/handoffs/<stage-name>.md. Survive cancellation for resume.
Team + Ralph Composition
When ralph modifier is present:
- Ralph creates PRD with stories
- Each story executed as a team pipeline
- Team verify/fix loop satisfies story acceptance criteria
- Ralph verifies story completion
- Next story begins
State Tracking
Track in .omc/state/team-state.json via omg_write_state/omg_read_state.
Resume and Cancel
- Resume: Restart from last non-terminal stage using state + handoffs
- Cancel:
/cancel→ graceful shutdown → cleanup → state preserved/cleared per policy
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/zereight/gitlab-mcp/team">View team on skillZs</a>