CONTINUITY: Security-Context Contracts for Composable LLM Agent Controls
The paper introduces CONTINUITY, a framework for verifiable composition of agent security controls, addressing security-context discontinuity where security-critical context may be dropped, widened, rebound, or reinterpreted across component boundaries. It models components with assume-guarantee contracts and carries authenticated security context using signed root grants, provenance commitments, role-bound transition receipts, bounded typed releases, transformation witnesses, and effect-bound execution permits. The framework formalizes end-to-end consequence integrity, requiring every realized external effect to be backed by a valid and current authorization witness linking principal, task, provenance, delegation, policy state, canonical action, and finality boundary. A reference verifier and deterministic cross-layer fault-injection suite covering 32 fault classes across four application domains are implemented.
CONTINUITY is a framework for verifiable composition of LLM agent security controls. It addresses security-context discontinuity, where individually correct security mechanisms fail to compose into end-to-end secure systems. The framework uses assume-guarantee contracts and authenticated security context across transitions, with mechanisms such as signed root grants, provenance commitments, role-bound transition receipts, bounded typed releases, transformation witnesses, and effect-bound execution permits. It formalizes end-to-end consequence integrity, requiring valid authorization witnesses for external effects. A reference verifier and fault-injection suite covering 32 fault classes across four domains are implemented.
The framework introduces assume-guarantee contracts for security components and authenticated context propagation via signed artifacts. It formalizes consequence integrity as a property linking authorization witnesses to external effects. The fault-injection suite with 32 fault classes provides a systematic method for evaluating composition failures.
As LLM agents are deployed in multi-component systems, security composition becomes critical. CONTINUITY offers a structured approach to ensure end-to-end security, potentially reducing integration risks for enterprises building agent pipelines.
For organizations deploying LLM agents, CONTINUITY could reduce security vulnerabilities arising from component composition, lowering risk and compliance costs. It may enable safer adoption of multi-vendor agent components.
Adoption of such frameworks may lead to standardized security contracts for agent components. Future work could extend to dynamic policy updates and cross-organizational agent interactions. Observable signals include open-source implementations or integration into agent orchestration platforms.