SHA-Pinning Bypass Exposes Four Major AI Coding Agents to Silent Code Injection

A zero-click remote code execution vulnerability called Plugin4Shell compromises four widely used AI coding agents by bypassing their plugin security pinning mechanism, potentially allowing attackers to inject malicious code into developer environments without user interaction. Researchers at Air Security discovered the flaw affects Claude Code, OpenAI Codex, GitHub Copilot, and Google's Gemini CLI—all of which rely on plugin marketplaces and SHA-pinning to verify plugin authenticity.