Agentic engineering practices changed from “you should commit everything yourself” to giving the agentic tools more autonomy for example in order to try out different things to implement a change. Small git commits are part of that toolset.

Here’s how to teach Codex CLI to self-identify as the git author.

Turns out that my Claude Code environment settings to make Claude identify itself as the git commit author are useful for audit trails – especially with the upcoming EU AI Act in August 2026, enabling agents to self-identify can be useful for engineering teams that tell their agents to create small, atomic git commits and then review branch diffs.

In ~/.codex/config.toml you can change the default git commit behavior from using your own personal user identity and have codex CLI use Codex <codex@openai.com>:

[shell_environment_policy]
inherit = "core"
set = {
GIT_AUTHOR_NAME = "Codex",
GIT_AUTHOR_EMAIL = "codex@openai.com",
GIT_COMMITTER_NAME = "Codex",
GIT_COMMITTER_EMAIL = "codex@openai.com"
}

For the environment inheritance rules, see: https://developers.openai.com/codex/config-advanced

If you use the pi coding agent, I shared the setup, too.