ResearchPCAS Policy Compiler for Secure Agentic Systems 48% to 93% CompliancearXiv·high signalXBlueskyLinkedInCopy linkCompiler-based deterministic policy enforcement improves agent compliance from 48% to 93% with zero violationsSourceSource pagearXiv↳ Follow the threadPolicy dependency / Stack layerSpecGuard Turns Speculative Decoding's Acceptance Rate Into a Free Runtime Backdoor DetectorarXivPolicy dependency / Stack layerOne in Seven Python Samples That Pass Bandit and Semgrep Still Carries a Runtime-Confirmed ExploitarXivPolicy dependency / Stack layerEvoSafeHarness searches policies and code together to build a per-model safety harness, cutting attack success from 45.6% to 10.0%arXiv (2609.05903)Policy dependency / Stack layerHolding Back Ready Agent Turns Instead of Releasing Them Eagerly Cuts P95 Workflow Latency up to 3.50xarXiv 2609.10964Policy dependency / Stack layerBenchShield raises reward-hacking detection in agent benchmarks from 23-94% to 77-100% full-chain recall, using a lifecycle model of reward eventsarXivPolicy dependency / Stack layerORCH Shows Org-Chart Design Beats Model Scale for 50-Agent Embodied Teams, Adding 63.97% to Mission ScorearXiv 2609.11737Policy dependency / ContrastIn a random draw of 400 MCP servers, only 48.8% complete a handshake, and 58.8% of tools omit safety annotationsarXiv 2609.10962Policy dependency / Update threadToxicRAG Poisons RAG With One 'Knowledge Update' Document per Target at 61-91% Attack SuccessarXiv