Loopic Agent Integration Skill

This skill helps AI assistants coordinate with the autonomous agent loops running in the orchestrator.

Overview

The n8n-infrastructure project uses autonomous loopic agents that run continuously in background loops:

• Monitoring Loop (30s): Health checks, anomaly detection
• Optimization Loop (5min): Performance tuning
• Predictive Loop (1min): Problem prediction
• Learning Loop (1h): Pattern learning
• Maintenance Loop (24h): Cleanup, backups

These agents are always running and handle most issues automatically. This skill helps you work with them, not against them.

---

Autonomous Agents in the System

Located in agents/orchestrator/src/agents/:

1. SupportAgent (support.agent.ts)

   • Auto-healing when health checks fail
   • Error rate monitoring
   • Automatic recovery attempts

2. DeploymentAgent (deployment.agent.ts)

   • Automated workflow deployment
   • Multi-tenant management
   • Validation and rollback

3. CustomizationAgent (customization.agent.ts)

   • Workflow personalization
   • Template management

4. MigrationAgent (migration.agent.ts)

   • Data migration support
   • Workflow import/export

---

When to Use This Skill

Check agent status BEFORE manual intervention when:

• User reports an issue that might already be handled
• Planning a system change that could conflict with auto-healing
• Investigating performance or reliability issues
• Wanting to understand what the system is doing automatically

---

Capabilities

1. Check Orchestrator Status

When to use: Before any manual intervention, to see if agents are handling it.

Instructions:

# Check if orchestrator is running
docker ps | Select-String "miconsul-orchestrator"

# If orchestrator is in this project (not all deployments have it running yet):
# View orchestrator logs
docker logs miconsul-orchestrator --tail 50

# Check for recent agent activity
docker logs miconsul-orchestrator --tail 100 | Select-String "SupportAgent|DeploymentAgent|CustomizationAgent"

Expected Output:

• If orchestrator running: Logs showing agent activities and loop executions
• If not running: Empty result (agents are code artifacts, not yet deployed in all environments)

What to do:

• ✅ If agents detected issue: Wait for auto-healing, monitor progress
• ⚠️ If no agent activity: Proceed with manual fix, but log what you did
• 🔴 If orchestrator not running: This is development/early deployment phase, proceed with manual operations

---

2. Check Event Bus Activity

When to use: To see what events the system is processing.

Instructions:

# If Redis is running (Event Bus):
docker ps | Select-String "redis"

# View Redis activity (if orchestrator is deployed):
docker logs miconsul-redis --tail 50

Event Types to Look For:

• HEALTH_CHECK_FAILED - SupportAgent will auto-heal
• ANOMALY_DETECTED - System is investigating
• OPTIMIZATION_SUGGESTED - System improving performance
• ERROR_RATE_INCREASING - Predictive alert

What to do:

• If events show agents are active → Let them work, monitor progress
• If no recent events → System is healthy or agents not deployed yet (check agents/orchestrator/dist)

---

3. Review Agent Loop Metrics

When to use: Understanding system health and autonomous performance.

Instructions:

# If orchestrator dashboard is available (future feature):
# curl http://localhost:3000/dashboard

# For now, check agent code to understand what loops do:
Get-ChildItem "agents\orchestrator\src\agents" -Filter "*.agent.ts" | Select-Object Name

# Read specific agent logic:
Get-Content "agents\orchestrator\src\agents\support.agent.ts"

Understanding the Loops (from docs/sistema_bucle_agentico.md):

Loop	Interval	Purpose
Monitoring	30s	Detect failures, anomalies
Optimization	5min	Improve performance
Predictive	1min	Predict future issues
Learning	1h	Learn patterns, best practices
Maintenance	24h	Cleanup, optimize DB

---

4. Coordinate with Auto-Healing

When to use: User reports issue, need to check if auto-healing is in progress.

Instructions:

# Check recent n8n logs for recovery attempts
docker logs n8n-v2 --tail 100 | Select-String "restart|recovery|auto-heal"

# Check if SupportAgent triggered recovery (if orchestrator running):
docker logs miconsul-orchestrator --tail 50 | Select-String "SupportAgent.*recovery"

# Check Docker restart count (indicates auto-healing attempts):
docker inspect n8n-v2 --format='{{.RestartCount}}'

Decision Tree:

Is there an error?
  ├─ Yes → Check auto-healing logs
  │   ├─ Auto-healing in progress? 
  │   │   ├─ Yes → Wait 2-5 minutes, monitor
  │   │   └─ No → Proceed with manual fix
  │   └─ Auto-healing failed?
  │       └─ Escalate or manual intervention
  └─ No → All good, no action needed

---

5. Understanding Tenant Context

When to use: Multi-tenant operations, understanding which client is affected.

Instructions:

# View tenant configuration (if exists):
Get-Content "agents\orchestrator\config\tenants.json" -ErrorAction SilentlyContinue | ConvertFrom-Json

# Or check environment for tenant info:
Get-Content ".env" | Select-String "TENANT"

Tenant Features (from agent code):

• autoHealing: Whether SupportAgent auto-restarts on failure
• autoOptimize: Whether OptimizationAgent applies fixes automatically
• autoUpdate: Whether system applies safe updates

Note: Current deployment may not have full multi-tenant setup yet. Check IMPLEMENTATION_STATUS.md for current phase.

---

6. Agent Deployment Status

When to use: Confirming which agents are actually running vs just code artifacts.

Instructions:

# Check if orchestrator container exists:
docker ps -a | Select-String "orchestrator"

# Check docker-compose for orchestrator service:
Get-Content "docker-compose.yml" | Select-String "orchestrator" -Context 5

# Check implementation status:
Get-Content "IMPLEMENTATION_STATUS.md" | Select-String "agent|orchestrator" -Context 2

Current State (as of 2026-01-16):

• Agents exist as code artifacts in agents/orchestrator/
• May not be deployed as running containers yet (check docker-compose.yml)
• Documentation exists in docs/sistema_bucle_agentico.md
• Follows a phased implementation approach

---

7. Manual Override Guidelines

When to use: You need to do manual operation despite autonomous agents.

Rules for Manual Intervention:

1. Always check first: Run diagnostics to see if agents are handling it
2. Wait for auto-healing: Give agents 5 minutes to resolve before manual fix
3. Log your actions: Document what you did so agents can learn from it
4. Coordinate with loops: Don't restart during a monitoring cycle (30s window)
5. Update Event Bus: If you fix something, emit an event so agents know

Example - Safe Manual Restart:

# 1. Check if auto-healing is active
docker logs n8n-v2 --tail 20 | Select-String "restart"

# 2. If no auto-healing in last 2 minutes, safe to proceed
# 3. Perform restart
docker-compose restart n8n

# 4. Monitor to ensure agents see the change
Start-Sleep -Seconds 30
.\monitor-n8n.ps1

---

8. Reading Agent Source Code

When to use: Understanding what agents can do automatically.

Key Files:

agents/orchestrator/src/
├── agents/
│   ├── support.agent.ts        # Auto-healing, error handling
│   ├── deployment.agent.ts     # Workflow deployment
│   ├── customization.agent.ts  # Personalization
│   └── migration.agent.ts      # Data migration
├── events/
│   ├── event-bus.service.ts    # Redis Pub/Sub
│   └── event-types.enum.ts     # All event types
└── loops/
    └── (future: monitoring.loop.ts, etc.)

Reading Agent Code:

# View SupportAgent capabilities:
Get-Content "agents\orchestrator\src\agents\support.agent.ts"

# See what events it listens to:
Get-Content "agents\orchestrator\src\agents\support.agent.ts" | Select-String "eventBus.on"

# Check available event types:
Get-Content "agents\orchestrator\src\events\event-types.enum.ts"

---

Workflow for AI Assistants

Standard Troubleshooting Flow

graph TD
    A[User Reports Issue] --> B{Check Orchestrator}
    B -->|Running| C[Check Agent Logs]
    B -->|Not Running| D[Use Manual n8n-management Skill]
    C --> E{Auto-Healing Active?}
    E -->|Yes| F[Wait 5min, Monitor]
    E -->|No| G[Check Event Bus]
    G --> H{Recent Events?}
    H -->|Yes| I[Agents are aware, wait]
    H -->|No| D
    F --> J{Issue Resolved?}
    J -->|Yes| K[Report Success]
    J -->|No| D

When to Choose Manual vs Autonomous

Scenario	Autonomous	Manual
n8n container down	✅ SupportAgent auto-heals	⚠️ If fails after 5min
Workflow JSON invalid	❌ Not auto-detected yet	✅ Use n8n-management
High error rate	✅ Monitoring Loop alerts	🔄 Support investigates
New workflow deploy	✅ DeploymentAgent	✅ Or manual via UI
Credential expiring	✅ Predictive Loop warns	✅ Manual renewal needed
Performance slow	✅ Optimization Loop fixes	⚠️ If persists

---

Expected Behavior

Healthy System

• Monitoring Loop running every 30s
• No critical events in last hour
• Auto-healing only triggered occasionally
• Optimization suggestions applied automatically

System Under Stress

• Frequent HEALTH_CHECK_FAILED events
• SupportAgent actively recovering
• Error rate increasing alerts
• Manual intervention may be needed

Development/Early Phase

• Orchestrator may not be deployed yet
• Agents are code-only (not running)
• Use manual skills from n8n-management
• Document issues for future agent training

---

Best Practices

1. Check Before Acting: Always check agent status before manual intervention
2. Respect Auto-Healing: Give agents time to resolve issues (5min window)
3. Coordinate Timing: Don't interrupt during monitoring loops (30s cycles)
4. Log Manual Actions: Document what you did for agent learning
5. Monitor Aftermath: Verify agents see your changes and adjust

---

Future Enhancements

As the orchestrator develops:

• Dashboard at http://localhost:3000/dashboard (planned)
• WebSocket for real-time event streaming
• Agent performance metrics
• ML-based prediction improvements

---

References

• docs/sistema_bucle_agentico.md - Full autonomous system design
• docs/agentes_mcp_resumen.md - Agent summaries and use cases
• agents/orchestrator/ - Agent source code
• IMPLEMENTATION_STATUS.md - Current deployment phase

---

Last Updated: 2026-01-16 Version: 1.0 Agent Deployment Phase: Check IMPLEMENTATION_STATUS.md for current status