Guardrails agents actually honor
Your team's Rust codebase has hard rules: all errors flow through a custom `AppError` type via `thiserror`, no `unwrap()` in non-test code, and all public functions are documented. You're giving an AI agent a task to add a new metrics-ingestion endpoint. How do you keep the agent inside these guardrails so its output doesn't need a convention-cleanup pass on review? What's the difference between a guardrail the agent will actually honor and one it'll quietly violate?
check_guardrails(lines: list[str]) → list[str][["/// Ingests one metrics batch.","pub fn ingest_metrics(body: String) -> Result<usize, AppError> {"," let parsed = parse_batch(body).unwrap();"," Ok(parsed)","}"]]out["3:unwrap_in_prod"][["pub fn ingest_metrics(body: String) -> Result<usize, AppError> {"," let parsed = parse_batch(body)?;"," Ok(parsed)","}"]]out["1:missing_doc"]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 & agentic: describe the solution in plain language (or narrate it) and the coach grades your approach.