Pipeline failure triage
A nightly data pipeline fails a few times a month, and each failure triggers a groggy debate about whether to backfill and who to warn downstream. Design a triage agent for pipeline failures: what it diagnoses, whether it may kick off a backfill itself, and how downstream consumers factor into its decisions. How would you evaluate it on past failures before giving it any authority, and pitch it to the data-platform lead in a few sentences.
triage_pipeline_failure(failure_class: str, is_idempotent: bool, prior_retries: int, downstream_consumers: int) → str["infra",true,0,3]out"retry_once"["schema_drift",true,0,2]out"notify_consumers_and_await_approval"["code",false,1,0]out"await_backfill_approval"Turn the fuzzy goal into a bounded agent task. Say what the agent reads, what it may do on its own versus draft for a human, and what it must never touch. Then earn the trust: how you’d verify it before rollout, and the pitch (the problem, what you built, why it matters).
Agent build: design the agent in plain language (or narrate it). The coach grades the decomposition, the guardrails, the verification plan, and the pitch.