The attacker optimizes to change this component's output. Three components qualify: the Domain LLM, which should emit the target string, and each Guard, which should score ALLOW. Each contributes a term to the objective.
The component sits between the suffix and a loss term, so the derivative has to travel through it. Nothing is trying to manipulate it. If it were a hard discrete branch the chain would break, and the suffix would be cut off from everything downstream.
The Coordinator builds an immutable Plan before touching external content. Flip anything the attacker controls and watch the plan stay identical. Flip a deployment switch and it changes.
The only input-dependent input to the plan, and it is a boolean for presence. The plan never reads the content itself.
Stage order is fixed by the architecture. The source comment is explicit: the plan does not read external_content, only its presence or absence.
Builds an immutable Plan before external content is touched and emits it as the first trace event, so an auditor can check that planned dispatch matched executed dispatch.
Conversation history with trust markers, the running aggregate trust score, and the active capability set for the session.
Four detectors and the Bayesian aggregation. This is Contribution 1 and the Coordinator's most substantive security work.
Set difference between the capabilities a response commits to exercising and those granted at admission. Deterministic, not a judgment call.
Compositional trust-role markers, Map-Reduce output assembly, and chat-template normalization across model families.
"Gradients flow through the Coordinator's routing decisions" implies there are meaningful decisions to flow through. In the delivered pipeline there are not: every branch tests configuration, and the one content-dependent input is a boolean for whether external content exists. The machinery is correct and general; it is simply not load-bearing today.
The plan appends the Input Guard before the multi-turn stage, while the handler runs the multi-turn pre-check first. If an auditor compares planned order against executed order rather than set membership, that difference would show up. It does not affect any reported result, but the auditability claim rests on the two matching.