docs: clarify ChatGPT full access behavior - #25806
Conversation
ChatGPT remote chats can continue requesting approvals even when Full access appears selected after reconnecting. Document how to select the mode, restart an affected task, and identify the upstream permission synchronization issue.
✅ Deploy Preview for docsdocker ready!
To edit notification comments on pull requests, go to your Netlify project configuration. |
docker-agent
left a comment
docker-agent
left a comment
There was a problem hiding this comment.
Assessment: 🟢 APPROVE
The PR adds a concise, well-scoped troubleshooting section clarifying how ChatGPT Full access works with Docker Sandboxes, including how to recover when approval prompts persist and a link to the upstream Codex issue. The new content is accurate, appropriately brief, avoids hedge words, marketing language, and meta-commentary, and matches the existing document's tone and scope. No missing redirects, broken markdown, front matter issues, or AI-pattern violations found.
docker-agent
left a comment
docker-agent
left a comment
There was a problem hiding this comment.
Assessment: 🟡 NEEDS ATTENTION
|
|
||
| ChatGPT controls Codex permissions separately from the Docker Sandbox. To run | ||
| commands without Codex approval prompts, select **Full access** from the | ||
| permissions menu for the remote chat. Enabling **Full access** under |
There was a problem hiding this comment.
[MEDIUM] Uses 'Enabling' instead of 'Turning on' for a UI toggle action
The sentence "Enabling Full access under Settings > General only adds the mode to the permissions menu" describes a user flipping a setting in the UI. The Docker style guide reserves "turn on"/"turn off" for UI toggles and settings a user flips in the interface.
Suggested rewording:
Turning on Full access under Settings > General only adds the mode to the permissions menu.
docker-agent
left a comment
docker-agent
left a comment
There was a problem hiding this comment.
Assessment: 🟡 NEEDS ATTENTION
|
|
||
| If the chat continues to request approval while showing **Full access**, update | ||
| the ChatGPT desktop app, then stop and restart the active task. Approval prompts | ||
| that return after reconnecting can indicate a |
There was a problem hiding this comment.
[MEDIUM] Weak qualifier "can indicate" leaves reader without actionable conclusion
The phrase "can indicate a Codex remote-permission synchronization issue" uses a weak qualifier that the style guide explicitly prohibits — avoid chains of "may," "can," and "depends" that leave the reader without a conclusion or next step.
Consider making this actionable:
If approval prompts return after reconnecting, see the [Codex remote-permission synchronization issue](https://github.com/openai/codex/issues/29054).
| commands without Codex approval prompts, select **Full access** from the | ||
| permissions menu for the remote chat. Enabling **Full access** under | ||
| **Settings > General** only adds the mode to the permissions menu. It doesn't | ||
| change an existing chat. |
There was a problem hiding this comment.
[LOW] Ambiguous pronoun "It" — repeat the noun for clarity
"It doesn't change an existing chat." uses an ambiguous pronoun. The style guide says to repeat the noun instead of relying on a pronoun to carry the reference across sentence boundaries.
Consider combining into one sentence:
Enabling **Full access** under **Settings > General** only adds the mode to the permissions menu and doesn't change an existing chat.
4 participants
Summary
Explain how Full access applies to ChatGPT remote chats and how to recover when approval prompts continue after reconnecting. Link to the upstream Codex permission synchronization issue for persistent cases.
@netlify /ai/sandboxes/integrations/chatgpt/
Preview: https://deploy-preview-25806--docsdocker.netlify.app/ai/sandboxes/integrations/chatgpt/
Generated by Codex