Blogs
Codex Sandbox Escapes Show Why Agent Guardrails Can’t Live Inside the Agent
Two patched OpenAI Codex vulnerabilities, Heapjack and Overpatch, exposed how coding agents can escape sandboxes and reach developer systems without approval prompts.
AI
A Semicolon in a Branch Name Was All It Took to Steal an AI Agent’s GitHub Token
AI coding agents don’t just suggest code anymore. Tools like OpenAI’s Codex spin up a real container, clone a real repository, and authenticate with a real GitHub credential to get the job done — which means every agent your team…