The doctrine that governs
every Foundry action.

The governance approach is not execution-layer guardrails. SOVEREIGN MIND shapes orientation before the agent acts — what the agent perceives, how it frames the situation, what doctrines apply. Guardrails come too late.

T1 Autonomous · T2 Supervised · T3 Commanded. Mode is a deployment-time mandate, not a runtime choice. T3 actions halt and hold for explicit human authority. Tier transitions require commander approval and are logged to CHRONICLE.

Before any agent action executes, the Gate node evaluates the action against the active ROE Card sequence: Charlie → Alpha → Bravo → Delta. Pass or halt. No partial execution. No exception path.

The L1 floor is immutable. No agent process — including the orchestrator — can modify it. It defines what no Foundry action will ever do, regardless of mandate. This is the limit beyond which intent does not extend.

Every agent execution is logged to CHRONICLE with full decision provenance. Inputs, gate evaluation, action taken, outcome, timing. The audit trail is the deliverable, not a by-product.

Each Foundry agent deployment carries a WARRANT certificate — the assurance framework that maps the agent's behaviour to the doctrine. WARRANT is the contract between the doctrine and the deployment.

Mandates are portable across Allied operations. Standing Orders translate between national doctrines. Multi-national engagement is a design primitive, not a retrofit.