SourcesOpenClaw Security Crisis 341 Malicious Skills CVE-2026-25253The Hacker News·high signalXBlueskyLinkedInCopy linkCVE-2026-25253 CVSS 8.8 one-click RCE. 42665 exposed instances. 341 of 2857 ClawHub skills malicious (12%). 335 traced to coordinated ClawHavoc operation.SourceSource pageThe Hacker News↳ Follow the threadStack layer / Threat patternCIPR: how you phrase a request changes whether a poisoned repo compromises your coding agentarXivStack layer / Threat patternThe runtime guard ships as an installable skill, and evolving it with MCTS beats pre-install vettingarXivStack layer / Threat patternA nine-stage lifecycle for agentic skills, from autonomous discovery through marketplace governancearXivPolicy dependency / Stack layerUpdate from Hell tests whether coding agents can survive dependency upgrades that break code silentlyarXivPolicy dependency / Threat patternJack Clark's read on the Hugging Face agent incident is that the agents coordinated, and that Ajeya Cotra called it 50% of the way to takeoverImport AI (Jack Clark)Policy dependency / Stack layerTop open source AI projects are auto-closing external PRs and letting agent "software factories" write a third of merged codeLatent SpacePolicy dependency / Stack layerUnsloth 0.1.805-beta doubles Qwen3.8-Flash-Next and GLM-5.3-Flash inference with MTP on by defaultGitHubStack layer / Threat patternContext Privilege Escalation Attacks Hit 12 Real Agent Harnesses, Including Claude Code and CodexarXiv 2609.01222