workflow-templates

DAG-Based Workflow Templates

Reusable workflow definitions imported from Agent33's orchestration engine. Each template is a JSON file describing a directed acyclic graph (DAG) of steps that an orchestrator can execute with dependency awareness, parallel execution, conditional branching, and retry logic.

What Are DAG-Based Workflows?

A DAG workflow models a task as a graph of steps where:

• Dependency-aware execution -- Steps declare which other steps they depend on via depends_on. The executor resolves the dependency graph and runs steps in topological order.
• Parallel execution -- Steps that share no dependencies run concurrently up to a configurable parallel_limit.
• Conditional branching -- Steps with action: "conditional" evaluate an expression and route execution through then or else sub-step arrays.
• Retry and timeout -- Each step can specify retry.max_attempts, retry.delay_seconds, and timeout_seconds for resilience.
• Input/output wiring -- Step outputs are referenced by downstream steps using template expressions like ${step_id.outputs.field} or {{ steps['step_id'].field }}.

Workflow JSON Schema

Every workflow file follows this top-level structure:

{
  "name": "workflow-name",
  "version": "1.0.0",
  "description": "What this workflow does",
  "triggers": {
    "manual": true,
    "on_event": ["event-name"],
    "on_change": ["glob/pattern/**"],
    "schedule": "cron-expression or null"
  },
  "inputs": {
    "param_name": {
      "type": "string|array|object",
      "description": "What this input is",
      "required": true,
      "default": null
    }
  },
  "outputs": {
    "result_name": {
      "type": "string|object|array",
      "description": "What this output contains"
    }
  },
  "steps": [ ... ],
  "execution": {
    "mode": "dependency-aware|sequential",
    "parallel_limit": 4,
    "continue_on_error": false,
    "fail_fast": true,
    "timeout_seconds": 300,
    "dry_run": false
  },
  "metadata": {
    "author": "agent-33",
    "created": "2026-01-30",
    "tags": ["tag1", "tag2"]
  }
}