OSSbenchflow-ai/skillsbench Agent Skills BenchmarkGitHub·high signalFirst standardized benchmark for agent skills. Key finding: curated skills boost 16.2% but self-generated skills provide zero benefit.SourceSource pageGitHub↳ Follow the threadPolicy dependency / Stack layerCynative Ships an Open-Source Deep-Research Security Agent With Default-Deny Write PermissionsIT Security NewsStack layer / ContrastDevRev launches Enterprise-Bench, an open benchmark for agents in messy enterprise conditionsITBriefStack layer / Threat patternMako: A 'Self-Evolving Agentic OS' That Writes New Exploits Against Live Targets and Hot-Loads Them Into ItselfarXivSame source domain / Semantic neighborGive long-horizon agents a message-gateway plus per-task sandboxed filesystem, as ByteDance DeerFlow does out of the boxGitHubStack layer / Threat pattern'Ghost Commit' Disclosure: PNG-Hidden Prompt Injection Makes Cursor and Anti-Gravity Exfiltrate .env Secretsawesome-ai-agent-attacks timelineStack layer / Update threadQwenPaw 2.0 Ships a Ground-Up 'Agent OS' Rewrite on AgentScope 2.0GitHubStack layer / Update threadShow HN: Jacquard — A Programming Language for AI-Written, Human-Reviewed CodeGitHub / Hacker NewsPolicy dependency / Update threadNetInjectBench: Naive Network-Ops Agents Take Unsafe Actions 82.5% of the Time Under Prompt InjectionarXiv