Pairs LLM-based causal knowledge extraction with an embedded deterministic validation layer, so that every extracted triplet is checked by inference rather than trusted on model output — the reliability core behind the .causal format. Presented at ICECET 2026 in Rome.