AgentsAWS Agentic AI Security Scoping Matrix — Four-Scope FrameworkAWS Security Blog·high signalXBlueskyLinkedInCopy linkFour scopes for agent security: read-only, propose changes, autonomous within bounds, self-initiating. Each maps to specific security controls.SourceSource pageAWS Security Blog↳ Follow the threadPolicy dependency / Stack layerRIPPLE: an edit confined to one prompt-policy segment changes downstream behavior, so replay candidate edits after previously accepted ones before persistingarXiv 2609.12127Stack layer / Threat patternAgentsDock Open-Sources an IDE That Collapses Termius, Cursor and Claude Code Into One DockAgentsDock (surfaced on Hacker News item 49678435)Policy dependency / Stack layerSGLang Hit With Unauthenticated Pickle RCE via /update_weights_from_tensor, the Fourth Critical Inference-Stack CVE in Four WeeksCERT Coordination CenterPolicy dependency / Stack layerSalesforce claims 7 billion Agentic Work Units delivered and gives Agentforce a runtime that pursues goals over days and weeksSalesforcePolicy dependency / Stack layerAgent Framework stops forwarding headers across redirects and revalidates file skill paths before useGitHubPolicy dependency / Stack layerPattern: four coding-agent CLIs shipped sandbox or trust work in the same weekGitHubStack layer / Threat patternTwo-gap framework recasts reward hacking and hallucination as symptoms of requirement and model gapsarXivStack layer / Threat patternControlled agentic CAD comparison: six unattended runs, 16 failures, 9 of which the tool never reportedModelRift
AWS Security Bloghigh signal