🔴 Chrome 151 fixes 6 critical memory flaws, 🎯 UNC6671 vishing gang hits hedge funds, ⚠️ Zero-privilege RCE in AI coding agents, ⚠️ TONTOU bypasses Spectre v2 mitigations, 🛡️ Cloudflare ships workerd sandbox fixes, 📋 Senate probes transnational scam coordination

The zero-privilege RCE in coding agents is the day's defining story. Malicious GitHub issues can now execute code in default configurations of Claude Code Action, Gemini CLI, and OpenAI Codex, a pattern demonstrated across more than 100 public repositories, including the vendors' own. The flaw is in the agent harness, permission logic, tool routing, and sandbox controls, not the model. Any automation that processes externally controlled text is in scope. Zero-Privilege RCE in Anthropic, Google, and OpenAI Coding Agents

The ChainDrop worm rotated its C2 domains through an Ethereum smart contract and began signing poisoned packages with genuine Sigstore provenance, which means valid provenance no longer implies a clean package. The opensearch-js path shows the worm still actively seeks new targets, with 453 public exfiltration repositories found across five accounts. ChainDrop npm Worm Rotates C2 via Ethereum Contract

Chrome 151 patches 41 flaws, including six critical memory-safety bugs in WebGL, Skia, and ANGLE; deploy it immediately on endpoints that touch untrusted sites. SpecterOps showed the WSUS patch channel can become a malware delivery channel when NTLM relay reaches the SUSDB database, and UNC6671's vishing campaigns against hedge funds add targeted MFA phishing to the financial-sector threat list.

Watchlist: agent harness permissions and PR review isolation, npm provenance verification, Chrome rollout, and WSUS authentication hardening.