Testing Governance Across the Agent Workflow - Public Research Brief

What evidence shows that a portfolio of governance tools preserves an authority boundary across multiple agents, tools and times? The source discussion includes approval, monitoring and shutdown, and treats governance as a technical and legal project. This brief extends that program by specifying the claims an acceptance exercise should be able to falsify. It does not attribute a deployment-only theory to the source. Declare the actors, grants, destination and protected acceptance event. Separate budget conservation, action scope, duplicate execution and useful work into distinct claims. Define what observation can support each. A policy describes permission; enforcement determines which effects can be stopped; observation establishes what occurred. A complete log is not itself prevention, and a system that refuses every request is not a useful purchasing workflow. This record is a substantive public research brief accompanying the full manuscript. The work is independent of AI Frontiers and is not presented as peer reviewed. Related author edition: https://highlandreflexivepress.net/research/safe-at-the-ribbon-cutting/#acquire AI tools assisted editorial preparation and formatting; Tony T. Nguyen is responsible for the research claims.