AgentHook: A Runtime Evidence Standard for Auditable AI Agent Governance

AI agents increasingly execute tool calls, delegate tasks, call models, write files, invoke APIs, and operate across enterprise systems. Existing governance approaches often rely on policy documents, prompt instructions, or application logs that do not provide a consistent account of what an agent was asked, what it intended, what tool it selected, what was executed, what returned, and which runtime controls were active at the moment of action. AgentHook is an open runtime evidence standard for AI agents. It defines a vendor-neutral event envelope, verified runtime contracts, canonical lifecycle event types, material tool-activity evidence, human decision records, incident signals, evidence-sealing records, conformance tiers, runtime attestation, and governance context metadata that allow agent runtimes to emit auditable evidence without binding adopters to a particular bus, policy engine, dashboard, or vendor.