This exercise is standard-library only by design: you hand-roll the normalization and validation mechanics that a schema library such as Pydantic automates, so the boundary rules stay visible. In production, reach for Pydantic — see the catalog's Pydantic documentation entry.

External payloads from webhooks, mobile clients, and LLM tool calls are untrusted. They frequently arrive with mixed camelCase and snake_case keys, dirty currency strings, or invalid enum states. A production boundary validator must sanitize valid variations and return structured errors without crashing.

python examples/pydantic-validation/verify.py starter --expect-failure
python examples/pydantic-validation/verify.py solution

The first command reports the expected failures from the naive starter. The second verifies the robust solution. Give TASK.md to a coding agent, ask it to fix only starter/validator.py, and run:

python examples/pydantic-validation/verify.py starter

See this example on flypython.com: the browsable contract, step-by-step walkthrough, and current review state.