Team Coordination Skill

Coordinate multi-agent workflows using teams, shared task queues, and inter-agent messaging.

What This Skill Does

• •Create and manage teams of AI agents
• •Distribute work via shared task queues
• •Enable communication between team members
• •Track task ownership and completion

When to Use This Skill

Use this skill when:

• •You need multiple agents to work on a complex task together
• •Work can be parallelized across specialist agents
• •Agents need to communicate findings or coordinate actions
• •You want to track who is working on what

Available Tools

This skill uses the following OpenCode tools (registered by opencode-teams plugin):

Team Management

• •spawn-team: Create a new team
• •discover-teams: List available teams to join
• •join-team: Join an existing team as a member
• •get-team-info: Get details about team members

Task Coordination

• •create-task: Add a task to the team's queue
• •get-tasks: View available or assigned tasks
• •claim-task: Claim a pending task to work on
• •update-task: Update task status or details

Communication

• •send-message: Send a direct message to a teammate
• •broadcast-message: Send a message to all team members
• •read-messages: Check messages from teammates

Example Workflows

Code Review Team

Leader Agent:
1. Use spawn-team to create "review-pr-456"
2. Use create-task to add:
   - Security review task
   - Performance review task
   - Style review task

Specialist Agents:
1. Use discover-teams to find "review-pr-456"
2. Use join-team to join as a member
3. Use get-tasks with filter {status: "pending"}
4. Use claim-task for their specialty
5. Complete review work
6. Use update-task to mark completed
7. Use send-message to report findings to leader

Leader Agent:
1. Use read-messages to collect all findings
2. Synthesize final review

Parallel Refactoring

Leader Agent:
1. Use spawn-team to create "refactor-services"
2. Break down work into service-specific tasks
3. Use create-task for each service

Worker Agents:
1. Use discover-teams to find work
2. Use join-team to join
3. Use get-tasks to see available work
4. Use claim-task for a service
5. Perform refactoring
6. Use broadcast-message to announce completion

Environment Variables

OpenCode sets these automatically:

• •OPENCODE_TEAM_NAME: Current team context
• •OPENCODE_AGENT_ID: Your unique agent identifier
• •OPENCODE_AGENT_NAME: Your display name
• •OPENCODE_AGENT_TYPE: Your role (leader, worker, specialist, etc.)

Best Practices

For Leaders:

• •Create specific, well-scoped tasks
• •Set appropriate task priorities
• •Monitor team progress via get-tasks
• •Synthesize results from multiple agents

For Workers:

• •Claim tasks that match your capabilities
• •Update task status regularly
• •Communicate blockers or questions
• •Share findings with relevant teammates

For Everyone:

• •Use descriptive team names (e.g., "review-pr-123" not "team1")
• •Keep messages concise and actionable
• •Clean up teams when work is complete (leader responsibility)

Error Handling

All tools throw descriptive errors:

• •Team already exists (when creating)
• •Team does not exist (when joining)
• •Task already claimed (when claiming)
• •Invalid team/task names

Check error messages and adjust accordingly.

Related Skills

This skill works well with:

• •Code analysis skills (for reviewing code)
• •Testing skills (for test coverage tasks)
• •Documentation skills (for doc generation tasks)

Tips

• •Start small: Create a team with 2-3 agents first
• •Use clear task descriptions
• •Establish communication patterns early
• •Leaders should periodically check task status
• •Workers should broadcast when they complete major milestones