Bringing Law to the Frontier
Vigil applies SwiftVector governance to OpenClaw with deterministic safety kernels. The case that behavioral guardrails are not governance — and what real enforcement looks like.
Agency Papers make the constitutional argument. Agent Statecraft is the operational layer — how you build the control plane, which enforcement patterns hold at scale, and what the evidence chain looks like in production. Technical depth. Real implementations. Verifiable claims.
The AI safety field has produced sophisticated arguments about why governance is necessary. What it has produced far less of is concrete architectural guidance on how to build governance systems that actually work in production — that survive contact with real agents, real tools, and real operational pressure.
Agent Statecraft closes that gap. Each article in this series is grounded in a working implementation — Vigil, FlightLaw, Chronicle Quest — and addresses a specific architectural question practitioners face when building governed AI systems.
The AgentVector framework →Articles build on each other. The Foundation track establishes the architectural argument. Implementation shows it across three domains. Patterns extract the repeatable methods. Field Guides are operational references you can work from directly.
The governance architecture doesn't change between domains. Only the Laws composed within it differ. These are the three domains where Agent Statecraft articles have been written or are underway.
Autonomous agents with shell, browser, and filesystem access. The canonical governance challenge: continuous action at machine speed, with a composition problem that per-action evaluation cannot catch.
Drone operations under FAA Part 107. The regulatory context makes the governance case explicit: deterministic safety boundaries are not optional when certification requires them.
Human authorship as a formal governance constraint. AI may draft prose; it cannot decide fate. Character mortality and plot resolution are human-only state transitions with audit trails.
Agent Statecraft is the operational counterpart to Agency Papers. Agency Papers establish why behavioral oversight cannot scale and what architectural governance must replace it. Agent Statecraft shows how to build that replacement — the control plane, the enforcement kernel, the evidence chain.
Read both in parallel: Agency Papers for the constitutional argument, Agent Statecraft for the engineering discipline.
Agency Papers →