ChatGPT Desktop's Codex agent forwards local SSH keys to remote sessions
A GitHub discussion on an OpenAI Codex pull request shows that ChatGPT Desktop's agent, when operating on a remote machine, can authenticate using the user's local SSH credentials via agent forwarding. This allows the remote-running agent to perform actions such as pushing commits to the user's private GitHub repositories using their own identity.
GoKawiil's interpretation of the reporting above, not reported fact.
SSH agent forwarding extends a user's local trust and credentials to remote environments, which could let a compromised or misbehaving remote session act with the full authority of the user's SSH key, including write access to private code. Security researchers often flag this forwarding pattern as risky because it blurs the boundary between local and remote trust, potentially exposing credentials to any process running on the remote host.
- ChatGPT Desktop's Codex agent can forward local SSH credentials to remote machines
- This could allow remote processes to push to private GitHub repos using the user's identity
- The behavior was surfaced via a GitHub pull request discussion, raising security concerns about credential exposure
Source: news.ycombinator.com, 2026-11-09
Published there as: “ChatGPT Desktop forwards your SSH agent to remote machines”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.