Formatting resolved, multi-system data into a coherent block — structured text, YAML/JSON, or a short narrative summary — suitable for injection into an LLM prompt, consistently enough that the same underlying data produces the same shape of block regardless of which task triggered it.