Agent Skills: Main Orchestrator

Main Orchestrator

Purpose

To decompose high-level requirements into concrete, testable tasks and coordinate execution across domain orchestrators. Maintains epic progress, resolves cross-domain conflicts, and ensures quality gates are met.

When to Use This Skill

Use this skill when:

• Receiving product requirements documents (PRDs)
• User submits feature requests spanning multiple domains
• Need to break down complex workflows into tasks
• Coordinating cross-domain dependencies (frontend + backend + data)
• Epic-level planning and progress tracking is required

How to Use This Skill

Step 1: Analyze Requirements

Parse incoming requirements to identify:

• Affected domains (frontend, backend, data, devops)
• Cross-domain dependencies
• Quality standards to apply
• Test requirements for each component

Step 2: Create Epic Structure

Generate an epic with decomposed tasks:

epic:
  id: "epic-{feature-name}"
  priority: "critical|high|medium|low"
  contexts: ["frontend", "backend", "data"]
  tasks:
    - task-001
    - task-002
  dependencies: {}
  success_criteria: "measurable outcomes"

Step 3: Decompose into Tasks

For each task, use this template:

task:
  id: "task-{number}-{feature}"
  epic: "parent-epic-id"
  context: "frontend|backend|data|devops"
  when: "Start conditions"
  who: "domain-orchestrator-skill"
  where: "Code locations"
  what: "Feature description"
  how: "standards/standard-name.md"
  goal: "Success criteria"
  check:
    valid: "Happy path scenarios"
    error: "Error handling tests"
    invalid: "Input validation tests"
    edge: "Boundary conditions"
    functional: "Business logic tests"
    visual: "UI tests (if applicable)"
    performance: "Load/speed tests"
    security: "Security validation"
  close: "Final state to record"

Step 4: Assign to Domain Orchestrators

Route tasks to appropriate orchestrators:

• frontend-orchestrator: UI/UX, components, state management
• backend-orchestrator: APIs, services, business logic
• data-orchestrator: ETL, pipelines, feature engineering
• test-orchestrator: Test strategy, coverage enforcement
• devops-orchestrator: Infrastructure, CI/CD, deployments

Step 5: Monitor and Coordinate

Track progress by:

• Reading operations.log for task completion events
• Updating ai-state/active/tasks.yaml with task status
• Resolving cross-domain conflicts
• Triggering human-docs generation on epic completion

Context Management

Read From

• ai-state/knowledge/patterns.md - Proven architectural patterns
• ai-state/knowledge/decisions.md - Past architecture decisions
• operations.log - Real-time event stream from all orchestrators
• standards/*.md - Quality standards to enforce

Write To

• ai-state/active/tasks.yaml - Task registry (living document)
• operations.log - Orchestration events (append-only)
• ai-state/knowledge/decisions.md - Strategic architectural choices

Communication Protocol

Listen for Events

{
  "event": "requirement.new",
  "source": "user|product-manager",
  "content": "requirement description"
}

Broadcast Events

{
  "event": "task.created",
  "orchestrator": "main",
  "task_id": "task-001",
  "assigned_to": "frontend-orchestrator",
  "dependencies": ["task-002"],
  "priority": "high"
}

Decision Criteria

When to Decompose vs Delegate

Decompose into multiple tasks when:

• Feature spans multiple domains (frontend + backend)
• Complex workflow with sequential steps
• Multiple skills need coordination
• Cross-domain dependencies exist

Delegate directly when:

• Single-domain feature (only frontend OR backend)
• Isolated component with no dependencies
• Single skill can handle end-to-end

Priority Assignment

1. Critical - Security issues, data loss, breaking changes
2. High - User-facing features, API changes, blockers
3. Medium - Performance improvements, refactoring
4. Low - Documentation, nice-to-have enhancements

Conflict Resolution

Backend-Frontend API Misalignment

Problem: Frontend needs API that backend hasn't built yet

Resolution:

1. Create mock API task for frontend
2. Update backend task with API contract
3. Add dependency: frontend depends on backend completion
4. Log decision in knowledge/decisions.md

Circular Dependencies

Problem: Task A needs B, Task B needs A

Resolution:

1. Identify shared requirement
2. Extract into new Task C
3. Make A and B both depend on C
4. Re-sequence task order

Resource Constraints

Problem: Multiple critical tasks, limited skill availability

Resolution:

1. Serialize tasks by business impact
2. Document delay reasoning in task notes
3. Communicate priority to stakeholders
4. Update timeline estimates

Success Metrics

Monitor these metrics:

• Task completion rate > 90%
• Cross-context conflicts < 5%
• Epic cycle time within estimates
• Test coverage > 85% per task
• Zero critical issues in production

Example Usage

Input: OAuth2 Login Feature

requirement: "Add OAuth2 login with Google and GitHub"

Output: Decomposed Epic

epic:
  id: "epic-oauth2"
  priority: "high"
  contexts: ["frontend", "backend", "data"]
  tasks:
    - task-001-oauth-ui:
        context: "frontend"
        who: "frontend-orchestrator"
        what: "Add Google and GitHub login buttons"
        how: "standards/frontend-standard.md"
        dependencies: []

    - task-002-oauth-api:
        context: "backend"
        who: "backend-orchestrator"
        what: "Implement OAuth2 flow with provider integration"
        how: "standards/backend-standard.md"
        dependencies: []

    - task-003-oauth-db:
        context: "data"
        who: "data-orchestrator"
        what: "Add OAuth tokens table and user linking"
        how: "standards/data-quality-standard.md"
        dependencies: []

    - task-004-oauth-tests:
        context: "test"
        who: "test-orchestrator"
        what: "E2E tests for OAuth flows"
        how: "standards/testing-requirements.md"
        dependencies: ["task-001", "task-002", "task-003"]

Anti-Patterns to Avoid

• Creating mega-tasks spanning multiple contexts without decomposition
• Skipping test requirements to "save time"
• Bypassing orchestrators with direct skill invocation
• Ignoring failed tests and proceeding to next task
• Document-first development (write code first, document after)
• Forgetting to log events in operations.log
• Not updating task status in ai-state/active/tasks.yaml

Integration Points

With Domain Orchestrators

frontend-orchestrator:

• UI/UX components
• State management
• Visual testing

backend-orchestrator:

• API development
• Business logic
• Service integration

data-orchestrator:

• Schema changes
• Data migrations
• Pipeline updates

test-orchestrator:

• Test strategy
• Coverage enforcement
• Quality gates

devops-orchestrator:

• Infrastructure provisioning
• CI/CD pipeline updates
• Deployment coordination

With Human-Docs Skill

Trigger documentation generation when:

• Epic completes successfully
• Breaking changes introduced
• Architecture decisions made

Provide to human-docs:

• Task summaries
• Architecture decisions
• API contract changes