Last released Oct 3, 2026
nab your agent's silent failures — verify LLM-agent outcomes against real system state, with self-heal. Zero dependencies.