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, governance context metadata, and transport guidance so agent runtimes can emit auditable evidence without binding adopters to a particular bus, policy engine, dashboard, or vendor. This draft preprint describes the problem, architecture, envelope model, runtime contract discovery, security considerations, interoperability model, versioning approach, and worked examples for AgentHook v0.1.

Draft preprint for review. This version refreshes the runtime contract discovery, tool-activity, human-decision, incident-signal, and evidence-sealing sections. The paper does not certify compliance with any law, regulation, or standard.