Question
You need a config loader for a Python service that reads settings from environment variables, applies defaults, coerces types (a port is an int, a feature flag is a bool, a timeout is a float), and fails loudly at startup if a required value is missing or malformed. You'll ask an AI agent to write it. Draft the prompt/spec with the constraints, edge cases, and acceptance criteria that make the agent get it right first try. What does a vague prompt ("parse my config from env") miss?
Treat the AI’s output as a draft to verify, not an answer to trust. Name the specific flaw and the input that triggers it, say how you’d catch it — tests, edge cases, reading critically — and how you’d re-prompt or decompose to get it right.
Vibe coding: describe the solution in plain language (or narrate it) and the coach grades your approach. Generating runnable code from your description is coming next.