Abstract
A system is proposed herein for governing artificial intelligence (AI) agents that perform network operations by interposing an evidence-capsule gate between deterministic network observations and an agent's memory and action authority. Structured evidence capsules bind operational facts and pass/fail assertions to topology scope, collection time, provenance, parser or schema version, freshness policy, and raw-evidence integrity information. A policy gate permits an agent to store trusted memory, cite current network truth, report successful operation, or execute or approve a network-changing action only when required capsules are fresh, in scope, integrity-verifiable, and passing. Missing, stale, conflicting, or tampered evidence causes blocking or deterministic re-observation. The Python Automation Test System (pyATS), other network testing and parsing tools, and a network simulation tool can provide practical evidence and tool interfaces, while the broader technique can be applied to other deterministic network evidence sources and can expose the same evidence ledger to dashboards, application programming interfaces (APIs), audit systems, continuous integration and continuous delivery (CI/CD) systems, change-management tools, and other agents.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Singh, Amit, "EVIDENCE-CAPSULE-GATED AGENT MEMORY AND ACTION CONTROL FOR NETWORK OPERATIONS", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/11772