The current external Gatekeeper white paper is Deterministic Execution Control for Enterprise AI Agents, External Release 1.0.
It supersedes the earlier July web edition. The paper moves the public architecture description to the enterprise-agent execution boundary: inference is proposal, authority exists at commit. It covers the canonical action envelope, deterministic rules as code, GREEN/YELLOW/RED authority outcomes, human review before commit, capability-scoped execution, Safety Receipts, ProofVault, post-execution state matching, deployment patterns, representative enterprise use cases, and the reference-candidate validation record.
The scope boundary is explicit. Reference-candidate benchmark evidence is not production certification. Production authorization remains topology-specific and requires site-specific deployment certification.