vulnerability
Malicious .git Configs Can Make Claude, Codex, Cursor, and Other AI Agents Run Attacker Code
HighCVSS 8.8
Summary
Manifold Security discovered eight security flaws in several AI coding agents – including Claude, Codex, and Hermes Agent – that allow attackers to execute arbitrary code on a user's machine. The vulnerability stems from the agents using Git's `core.fsmonitor` setting to run commands from a repository's `.git/config` file, bypassing sandboxes and requiring no user approval. While some fixes have been released, several agents remain vulnerable, requiring users to take proactive steps to mitigate the risk.
Summary written automatically in our own words from the original article, which belongs to its publisher and remains the reference. It may contain errors. Sources & data
