Security researchers have uncovered a critical command-injection vulnerability in OpenAI’s Codex CLI, identified as CVE-2025-61260, which allows attackers to execute arbitrary commands on a developer’s machine without any interaction. The flaw stems from how Codex CLI automatically loads Model Context Protocol (MCP) server entries from local project configuration files. When a developer runs Codex inside a repository, these entries execute silently without prompts, validation, or approval, turning ordinary configuration files into a powerful execution channel for malicious code. The attack works by modifying repository files that developers typically trust. An attacker with write access or the ability to submit a pull request can place a malicious .env file to redirect Codex’s configuration directory, along with a crafted config.toml containing harmful MCP entries. Once a developer clones the repository and runs Codex, the malicious commands execute automatically. This enables reverse shells, credential theft, file manipulation, and persistent backdoors. The risk extends further into supply-chain environments: continuous integration systems, build pipelines, or open-source templates that run Codex can also become compromised, propagating malicious code to downstream users. To mitigate risk, organizations should temporarily avoid running Codex CLI on untrusted repositories, review local configuration loading behavior, and audit existing projects for suspicious MCP entries. Developers should restrict repository write permissions, enforce strict PR reviews, and enable monitoring for unusual environment variable changes or unexpected .codex directories. Until a patched version is released, teams are strongly advised to treat Codex CLI as potentially dangerous in shared or open-source environments and apply standard supply-chain security controls such as code signing, environment isolation, and CI/CD hardening.
Researchers have unveiled a critical flaw in the .NET Framework, dubbed SOAPwn, that can be exploited to force enterprise applications to perform arbitrary file writes and achieve ...
Google has released an emergency security update for Chrome, addressing a critical zero-day vulnerability that is actively being exploited by attackers. The update, which brings Ch...
A novel Android malware strain, dubbed DroidLock, has been identified that locks victims’ screens and extorts ransom payments while threatening to destroy their data. Discovered ...