Temporal AI Agents - Claude/Codex Skills Interface

Overview

The Temporal AI Agents system provides a comprehensive set of skills for managing and orchestrating AI agent workflows. This skill interface allows Claude and Codex to interact with the system through structured tool calls, enabling automated compliance checking, security analysis, cost optimization, and workflow management.

System Architecture

The system consists of:

• Temporal Workflow Engine: Durable orchestration of AI agent workflows
• Multi-Agent Framework: Specialized agents for compliance, security, and cost analysis
• REST APIs: Programmatic access to workflow management
• MCP Server: Standardized agent-tool communication protocol
• Infrastructure Emulator: Safe simulation environment for cloud resources

Available Skills

1. Compliance Management Skills

start_compliance_check

Starts a compliance check workflow for a target resource.

Parameters:

• targetResource (string, required): The resource to check (e.g., "vm-web-server-001")
• complianceType (string, optional): Type of compliance check - "SOC2", "GDPR", "HIPAA", "full-scan" (default: "full-scan")
• priority (string, optional): Priority level - "low", "normal", "high", "critical" (default: "normal")

Returns:

• workflowId (string): Unique identifier for the started workflow
• status (string): Initial status ("started")
• targetResource (string): Resource being checked
• complianceType (string): Type of compliance check
• startedAt (string): ISO timestamp when workflow started

Example:

{
  "targetResource": "vm-web-server-001",
  "complianceType": "SOC2",
  "priority": "high"
}

get_compliance_status

Retrieves the current status of a compliance workflow.

Parameters:

• workflowId (string, required): The workflow ID to check

Returns:

• workflowId (string): Workflow identifier
• status (string): Current status ("running", "completed", "failed")
• complianceScore (number): Compliance score (0-100)
• issuesFound (number): Number of issues identified
• recommendations (array): List of remediation recommendations
• approved (boolean): Whether the resource passed compliance checks

2. Security Analysis Skills

start_security_scan

Initiates a security analysis workflow.

Parameters:

• targetResource (string, required): Resource to scan
• scanType (string, optional): Scan type - "vulnerability", "malware", "configuration", "full" (default: "full")
• priority (string, optional): Priority level (default: "normal")

Returns:

• workflowId (string): Workflow identifier
• status (string): Initial status
• targetResource (string): Resource being scanned
• scanType (string): Type of security scan
• startedAt (string): Start timestamp

get_security_report

Retrieves security analysis results.

Parameters:

• workflowId (string, required): Security workflow ID

Returns:

• vulnerabilities (array): List of identified vulnerabilities
• riskLevel (string): Overall risk assessment ("low", "medium", "high", "critical")
• recommendations (array): Security remediation steps
• scanCoverage (number): Percentage of resource scanned

3. Cost Optimization Skills

start_cost_analysis

Begins cost optimization analysis.

Parameters:

• targetResource (string, required): Resource to analyze
• analysisType (string, optional): Analysis type - "usage", "optimization", "forecast", "full" (default: "full")
• timeframe (string, optional): Analysis period - "7d", "30d", "90d" (default: "30d")

Returns:

• workflowId (string): Analysis workflow ID
• targetResource (string): Resource analyzed
• analysisType (string): Type of cost analysis
• timeframe (string): Analysis period

get_cost_recommendations

Retrieves cost optimization recommendations.

Parameters:

• workflowId (string, required): Cost analysis workflow ID

Returns:

• currentCost (number): Current monthly cost
• projectedSavings (number): Potential monthly savings
• recommendations (array): Specific cost optimization actions
• roi (number): Return on investment percentage

4. Workflow Management Skills

list_active_workflows

Lists all currently running workflows.

Parameters:

• workflowType (string, optional): Filter by type ("compliance", "security", "cost-analysis")

Returns:

• workflows (array): List of active workflows with IDs, types, and start times

get_workflow_details

Gets detailed information about a specific workflow.

Parameters:

• workflowId (string, required): Workflow identifier

Returns:

• workflowId (string): Workflow ID
• type (string): Workflow type
• status (string): Current status
• startTime (string): When workflow started
• endTime (string, optional): When workflow completed
• progress (number): Completion percentage (0-100)
• result (object): Workflow results (varies by type)

cancel_workflow

Cancels a running workflow.

Parameters:

• workflowId (string, required): Workflow to cancel
• reason (string, optional): Reason for cancellation

Returns:

• success (boolean): Whether cancellation was successful
• message (string): Confirmation message

5. Infrastructure Discovery Skills

discover_resources

Discovers available infrastructure resources.

Parameters:

• resourceType (string, optional): Filter by type - "vm", "database", "storage", "network", "all" (default: "all")
• environment (string, optional): Filter by environment - "dev", "staging", "prod", "all" (default: "all")

Returns:

• resources (array): List of discovered resources with metadata
• totalCount (number): Total number of resources found

get_resource_details

Gets detailed information about a specific resource.

Parameters:

• resourceId (string, required): Resource identifier

Returns:

• resourceId (string): Resource identifier
• type (string): Resource type
• environment (string): Deployment environment
• status (string): Current operational status
• configuration (object): Resource configuration details

6. Human-in-the-Loop Skills

request_human_review

Initiates a human review workflow for agent decisions.

Parameters:

• workflowId (string, required): Workflow requiring human review
• reviewType (string, required): Type of review needed
• priority (string, optional): Review priority ("low", "normal", "high", "critical")
• context (object, optional): Additional context for reviewers

Returns:

• reviewId (string): Human review task identifier
• assignedTo (string): Team or person assigned for review
• estimatedCompletion (string): Expected completion time

get_review_status

Checks the status of a human review task.

Parameters:

• reviewId (string, required): Review task identifier

Returns:

• status (string): Review status ("pending", "in-progress", "completed")
• decision (string, optional): Final decision if completed
• comments (string, optional): Review comments
• reviewedBy (string, optional): Who performed the review

Usage Examples

Compliance Automation Workflow

// Start compliance check
const complianceWorkflow = await start_compliance_check({
  targetResource: "prod-web-server-001",
  complianceType: "SOC2",
  priority: "high"
});

// Monitor progress
let status;
do {
  status = await get_workflow_details({ workflowId: complianceWorkflow.workflowId });
  await new Promise(resolve => setTimeout(resolve, 5000)); // Wait 5 seconds
} while (status.status === "running");

// Get results
if (status.status === "completed") {
  console.log(`Compliance Score: ${status.result.complianceScore}%`);
  console.log(`Issues Found: ${status.result.issuesFound}`);
  if (!status.result.approved) {
    console.log("Recommendations:", status.result.recommendations);
  }
}

Security Incident Response

// Detect and respond to security incident
const securityScan = await start_security_scan({
  targetResource: "compromised-server-001",
  scanType: "full",
  priority: "critical"
});

// Wait for completion and get report
const report = await get_security_report({ workflowId: securityScan.workflowId });

// Take automated actions based on findings
if (report.riskLevel === "critical") {
  // Trigger emergency response workflows
  await request_human_review({
    workflowId: securityScan.workflowId,
    reviewType: "security-incident",
    priority: "critical"
  });
}

Cost Optimization Analysis

// Monthly cost review
const costAnalysis = await start_cost_analysis({
  targetResource: "all-production-resources",
  analysisType: "full",
  timeframe: "30d"
});

// Get optimization recommendations
const recommendations = await get_cost_recommendations({
  workflowId: costAnalysis.workflowId
});

console.log(`Current Cost: $${recommendations.currentCost}/month`);
console.log(`Potential Savings: $${recommendations.projectedSavings}/month`);

// Implement high-ROI recommendations automatically
recommendations.recommendations
  .filter(rec => rec.roi > 50)
  .forEach(rec => {
    console.log(`High ROI Recommendation: ${rec.description}`);
    // Trigger implementation workflow
  });

Integration Guidelines

Authentication

All skill calls require API key authentication:

Authorization: Bearer YOUR_API_KEY

Rate Limiting

• 100 requests per minute per API key
• 1000 requests per hour per API key
• Batch operations encouraged for bulk actions

Error Handling

All skills return consistent error responses:

{
  "error": {
    "code": "VALIDATION_ERROR",
    "message": "Invalid workflow ID format",
    "details": { "field": "workflowId", "expected": "uuid-format" }
  }
}

Timeouts

• Workflow operations: 15 minutes
• Status queries: 30 seconds
• Resource discovery: 2 minutes

System Requirements

• Backend API: http://localhost:8081 (configurable)
• MCP Server: localhost:8082 (stdio/websocket/http)
• Authentication: API key required
• Data Format: JSON for all requests/responses

Monitoring and Logging

All skill executions are logged with:

• Timestamp and duration
• User/API key identification
• Success/failure status
• Performance metrics
• Audit trail for compliance

This skill interface enables Claude and Codex to seamlessly integrate with the Temporal AI Agents system for automated infrastructure management, compliance monitoring, and intelligent workflow orchestration.