Containment Failures in AI Coding Agents
Six vulnerabilities in Codex and Gemini CLI reveal how startup, sandboxing, configuration, and approval checks in coding agents can fail.
AI Security
Research, writeups, and notes on things I find interesting.