Does not adhere to gpg sign commit rule
What version of Codex is running?
0.53.0
What subscription do you have?
Pro
Which model were you using?
gpt-5-codex high, gpt-5 high
What platform is your computer?
Darwin 25.1.0 arm64 arm
What issue are you seeing?
I have noticed this on every other session where I ask the model to commit the changes. I have configured the git commit using GPG code sign, turned on and using 1Password to maintain the GPG key. So, every other git commit method that I use outside this CLI asks for authentication if 1Password is logged in and then commits it with the GPG sign. But Codex alone here doesn't ask for that permission, and it fails a couple of times. So it turns off GPG sign for this local repository and then commits the changes.
<img width="1313" height="663" alt="Image" src="https://github.com/user-attachments/assets/db5d81cf-303b-4926-bf94-63b39cfd6557" />
What steps can reproduce the bug?
019a4f2b-4239-7333-9226-d5f32ba9ab01
Enable GPG code-sign and have 1password, locked or unlocked, doesn't matter. Ask the model to commit the changes. It doesn't adhere to the standard methods and turns it off.
What is the expected behavior?
_No response_
Additional information
_No response_
4 Comments
Are you running with
--dangerously-bypass-approvals-and-sandbox/--yoloor any other changes to the default sandbox?It's possible the default sandbox is getting in the way of accessing the resources required to GPG sign. You can test this with:
Also, we make the
.gitfolder within a repo read-only by default on macOS today.I do have
sandbox_mode = "danger-full-accessand this line on AGENTS.md<details>
<summary>Full Config</summary>
Core (root defaults)
notify = ["/Users/cyron/.codex/notify.sh"]
model = "gpt-5"
approval_policy = "on-request" # untrusted|on-failure|on-request|never
sandbox_mode = "workspace-write" # read-only|workspace-write|danger-full-access
profile = "yolo"
tool_output_token_limit = 25000
model_reasoning_summary_format = "experimental"
Extra settings that only apply when
sandbox = "workspace-write".[sandbox_workspace_write]
Allow the command being run inside the sandbox to make outbound network
requests. Disabled by default.
network_access = true
UI & notifications (can be toggled by installer)
[tui]
Desktop notifications from the TUI: boolean or filtered list. Default: false
Examples: true | ["agent-turn-complete", "approval-requested"]
notifications = true
Centralized feature flags — booleans only
show_raw_agent_reasoning = true
hide_agent_reasoning = false
[features]
web_search_request = true
unified_exec = false
streamable_shell = false
rmcp_client = true
apply_patch_freeform = false
view_image_tool = true
experimental_sandbox_command_assessment = false
ghost_commit = false
enable_experimental_windows_sandbox = false
---- Profiles (override only what differs from root) -----------------------
[profiles.balanced]
approval_policy = "on-request"
sandbox_mode = "workspace-write"
[profiles.balanced.features]
web_search_request = true
[profiles.safe]
approval_policy = "on-failure"
sandbox_mode = "workspace-write"
[profiles.safe.features]
web_search_request = false
[profiles.minimal]
model_reasoning_effort = "minimal"
[profiles.minimal.features]
web_search_request = false
[profiles.yolo]
approval_policy = "never"
sandbox_mode = "danger-full-access"
model = "gpt-5.1-codex-max"
model_reasoning_effort = "high"
exec_timeout_ms = 900000
[profiles.yolo.features]
web_search_request = true
</details>
Reviewing old (possibly stale and no-longer-relevant) bug reports. Are you still seeing this with the latest version of Codex?
This issue is very specific, and it hasn't been submitted by any other users, so it hasn't been a high priority for us to look at. You may want to do some additional investigation and diagnosis on your own. You could clone the codex code base and compile it locally, then pair with codex to look for the issue.
Apologies for the late reply. This is no longer an issue.