Abstract page for arXiv paper 2608.16891: Runtime Governance for Agentic AI: Action-Boundary Control with Trusted Provenance and Fail-Closed Execution