Action Agent Details

Action Agent Details

Action Agents execute remediation actions instructed by the orchestrator, serving as the hands of the platform.

Core Functionality

Action Agents receive instructions from the orchestrator and execute them through secure, sandboxed plugins.

Instruction Processing

Inbound: Polls for instruction objects from the orchestrator

{
  "id": "uuid",
  "plugin_id": "uuid",
  "plugin_configuration": {},
  "input_data": {},
  "timeout_seconds": 300,
  "max_retries": 3
}

Outbound: Reports execution result back to the orchestrator

{
  "status": "completed|failed|timeout",
  "result": {},
  "error_message": "string",
  "duration_ms": 1500,
  "resource_usage": {}
}

Execution Flow

1. Poll orchestrator API for pending instructions
2. Download and validate required action plugin
3. Execute plugin in isolated sandbox environment
4. Report execution results back to orchestrator

Plugin Architecture

Action Agents use Go-based plugins that execute in secure sandbox environments.

Plugin Lifecycle

graph TD
    A[Plugin Registry] --> B[Download Plugin]
    B --> C[Validate Plugin]
    C --> D[Load Plugin]
    D --> E[Initialize Plugin]
    E --> F[Execute Plugin]
    F --> G[Report Results]
    G --> H[Unload Plugin]

Security Controls

• Resource Limits: Memory and CPU usage strictly controlled
• Filesystem Access: Limited to allowed directories only
• Network Access: Restricted based on plugin requirements
• System Calls: Whitelist of allowed operations

Configuration

Action Agents are configured via YAML with plugin repositories, execution parameters, and security settings.

agent:
  type: action
  id: action-001
  orchestrator_url: https://agents.stavily.com
  auth_token: ${AGENT_TOKEN}
  polling_interval: 5s

plugins:
  repositories:
    - name: system-commands
      url: https://github.com/stavily/system-commands
      version: v1.0.0

execution:
  timeout: 300s
  retries: 3

security:
  plugin_sandbox: true
  resource_limits:
    memory: 512MB
    cpu: 1.0

Monitoring

Action Agents provide structured logging and metrics for observability.

{
  "timestamp": "2025-01-15T10:30:00Z",
  "level": "INFO",
  "component": "action-agent",
  "agent_id": "action-001",
  "event": "plugin_executed",
  "plugin": "restart-service-v1.0.0",
  "duration_ms": 1500,
  "status": "completed"
}