fail-loud
agentvitals — a small, dependency-free reliability harness for scheduled/autonomous AI agents: persist state across runs, prove liveness (not just freshness), fail loud.
