What problem does it solve? Orchestrating a multi-step agent pipeline with hardcoded step logic is brittle and hard to resume after interruptions. This Skill replaces hardcoded routing with a declarative, machine-readable workflow graph (DAG) that the Orchestrator reads to determine the next step, handle human approval gates, and resume from checkpoints. ## Core Features & Use Cases - DAG-Based Step Routing: Loads templates/workflow-graph.json and follows on_complete, on_skip, and on_fail edges to route between agent steps, gates, validations, and fan-out nodes. - Conditional IaC Branching: Routes Steps 4-6 to Bicep (*-b) or Terraform (*-t) tracks based on the decisions.iac_tool value captured in requirements. - Gate and Fan-Out Handling: Blocks downstream execution at human approval gates and dispatches parallel sub-steps (e.g., Step 7 documentation generation) concurrently. - Use Case: After a user approves the architecture gate, the Orchestrator runs apex-recall show <project> --json, finds the current node in the graph, and routes to governance discovery or the correct IaC planning branch without any hardcoded logic. ## Quick Start Ask the orchestrator to load the workflow graph and resume my project from its current step using apex-recall session state.
