merge-authority

The method's canonical merge-authority contract: for each loop model, who executes a merge, the verdict literals, the complete human-hold set and how the safe and boundary classes relate; and the one block a project brief declares instead of restating it (its loop model and its own boundary triggers). Use when classifying a merge, writing or reading a gate verdict, deciding who merges, or writing a project brief's merge section. A project may add boundary triggers; it may not change the executor, the verdicts or the holds.