workflow-runner

You can run YAML workflows directly inside Claude or Cursor by providing the YAML file, using the active LLM session as the execution engine without needing any API keys.

Executing multi-role AI agent tasks without API keys is done by resolving roles, rendering task templates, and coordinating sequential or parallel steps using the current chat session as the execution engine.

YAML parsing and role resolution work by validating the user-provided YAML workflow file, automatically resolving role directories, and rendering task templates to coordinate multi-step execution across agents.

Yes, you can automate sequential and parallel workflow steps directly in your chat by providing a YAML workflow file or describing the multi-role task to the current LLM session.

You should use YAML workflow orchestration when you need structured multi-role collaboration, context management across steps, and preserved step outputs for audit and reuse rather than manual prompting.

Yes, the workflow runner supports preserving step outputs for audit by enforcing YAML parsing, managing context across execution steps, and producing a final summary alongside any generated artifacts.