Organize reusable prompt workflows with POML

Use POML when an operator needs prompt workflows that are too important or complex to leave as ad hoc chat text. The operator writes prompt assets in POML markup with role, task, examples, data inputs, images, tables, output format rules, variables, loops, conditionals, and presentation styles, then renders or tests them through the VS Code extension, Python package, or JavaScript package. Invoke this instead of ordinary manual prompting when prompts must be versioned, reused across applications, parameterized with external data, inspected in code review, or rendered consistently for different LLM providers. The scope boundary is prompt orchestration and prompt asset maintenance: structure, template, render, preview, and test repeatable prompts. It is not a generic framework listing or model provider card; the workflow is specifically about maintaining prompts as source-controlled operator assets with deterministic inputs and reviewable outputs. Keep each POML file tied to a named task, expected inputs, and output contract so agents can reuse it safely.