What problem does it solve? Running agency-orchestrator YAML workflows normally requires a CLI setup with API keys. This Skill lets you execute multi-role orchestration workflows directly inside Claude Code, OpenClaw, or Cursor, using the current session's LLM as the execution engine. ## Core Features & Use Cases - YAML Workflow Execution: Parses workflow files with inputs, steps, roles, and dependency graphs, then executes them layer by layer with topological ordering. - Multi-Role Orchestration: Reads agent role definitions from agency-agents-zh and impersonates each role with its expertise and communication style, running parallel steps via sub-agents. - Structured Output Persistence: Saves per-step outputs, a final summary, and metadata.json into a dated .ao-output directory. - Use Case: A user provides a story-creation.yaml workflow where a narrator, psychologist, and editor collaborate; the Skill runs each role in dependency order and delivers the final story plus all intermediate outputs. ## Quick Start Run the workflow defined in workflows/story-creation.yaml using the agency-agents-zh role definitions.
