It Agent Coordination

IT Agent Coordination

Overview

This skill covers multi-agent workflow orchestration for IT operations. It encompasses workflow definition, agent-to-agent handoffs with structured data, error recovery patterns, workflow monitoring, and agent lifecycle management. Primary agent: 02050-006 AI Orchestrator. Supporting skills: All domainforge_ai IT skills.

Triggers

• Multi-agent workflow needs to be initiated
• Agent execution failure requires recovery
• New workflow needs to be orchestrated
• Existing workflow requires monitoring or intervention
• Agent lifecycle event detected (start, complete, fail, timeout)

Prerequisites

• Agents defined and available in agent registry
• Workflow schema and handoff definitions established
• Error handling and recovery patterns defined
• Monitoring and alerting configured
• Agent communication channels established

Steps

Step 1: Workflow Definition

• Define workflow stages and agent sequence
• Define data schema for each handoff
• Define success criteria for each stage
• Define failure handling (retry, alternative, escalation)
• Define timeout thresholds for each stage

Step 2: Agent Assignment & Validation

• Verify all required agents are available
• Confirm agent capabilities match workflow requirements
• Assign primary and backup agents for each stage
• Validate agent environment connectivity
• Confirm agent authentication and authorization

Step 3: Handoff Configuration

• Define data format for each agent handoff (structured JSON, not free text)
• Define validation rules for handoff data
• Define data transformation rules between incompatible formats
• Set up handoff monitoring and logging

Step 4: Workflow Initiation

• Trigger workflow with initial data
• Verify first agent receives correct input
• Monitor first agent execution progress
• Validate first agent output against expected schema

Step 5: Handoff Execution

• Pass validated output from Agent A to Agent B
• Verify Agent B acknowledges receipt and begins processing
• Monitor execution time against thresholds
• Check output quality against success criteria

Step 6: Error Handling & Recovery

• Detect agent execution failure (error, timeout, invalid output)
• Apply error recovery strategy (retry, alternative path, human escalation)
• Log recovery action with reasoning
• Resume workflow from last valid checkpoint

Step 7: Workflow Monitoring

• Track workflow progress through stages
• Monitor stage durations against baselines
• Detect stalled workflows
• Alert on workflow failures or SLA breaches

Step 8: Workflow Completion

• Validate final output against workflow success criteria
• Generate workflow execution summary
• Log audit trail for all actions and decisions
• Update workflow analytics and metrics

Success Criteria

• All workflow stages complete successfully
• Data handoffs use validated structured JSON (no free text between agents)
• Error recovery succeeds for all recoverable failures
• Workflow completes within defined SLA
• Complete audit trail maintained
• All stages logged with execution metrics

Common Pitfalls

1. Unvalidated handoffs: Always validate handoff data before passing to next agent
2. Missing backups: Define alternative paths for critical agent failures
3. Timeout storms: Set appropriate timeouts — too short causes cascading failures, too long wastes resources
4. Free text handoffs: Use structured JSON — free text loses schema guarantees
5. No monitoring: Always monitor workflow progress and stages
6. Orphaned workflows: Define cleanup for stuck or failed workflows

Cross-References

• it-error-discovery-classification/SKILL.md — Error classification for agent failures
• it-log-analysis-monitoring/SKILL.md — Workflow execution log monitoring

Usage

Apply this skill when setting up or monitoring multi-agent workflows, handling agent failures during workflow execution, or designing new orchestrated workflows.

Metrics

• Workflow Completion Rate: 95%+ of workflows complete successfully
• Handoff Accuracy: 100% of handoffs use validated structured JSON
• Error Recovery Rate: 90%+ of recoverable failures resolved without human intervention
• SLA Compliance: 95%+ of workflows complete within SLA
• Audit Trail: 100% of workflow actions logged with source, timestamp, and decision