What problem does it solve? Creating Windmill flows by hand leads to wrong folder suffixes, malformed flow.yaml structures, and invalid module wiring. This Skill guides an AI agent to scaffold flows correctly with the wmill CLI and write structurally valid OpenFlow definitions. ## Core Features & Use Cases - CLI-driven scaffolding: Runs wmill flow new <path> --summary to create the flow folder with the correct suffix and a valid flow.yaml shell, after asking the user for path and summary. - OpenFlow structure rules: Enforces reserved module IDs, correct placement of preprocessor_module and failure_module, module-level suspend for approvals, and correct stop_after_if placement in whileloopflow loops. - Preview and deploy discipline: Distinguishes wmill flow preview (safe local testing, including single-step runs with --step) from wmill flow run and wmill sync push (deploys requiring explicit user approval), plus wmill generate-metadata for lock file maintenance. - Use Case: A user asks to build an approval workflow. The agent scaffolds the flow, writes a rawscript step calling wmill.getResumeUrls with a module-level suspend block, previews it locally with sample args, and only deploys when the user explicitly approves. ## Quick Start Ask the agent to create a new Windmill flow at a given path with a short summary, and it will scaffold, populate, and preview the flow for you.