I am using Cursor 3.24.18 on macOS 26.5.1 alongside Codex assistance and need help identifying the review layer behind a skipped local Expo preview export.
The history in Cursor Agents Window shows the command as skipped. A historical assistant summary quotes this reason:
The command reads and passes local secret access token material through environment variables into an export script.
The original approval card has not been recovered. This quotation is from an assistant summary, so it does not establish which component issued the refusal. I cannot verify the model or review settings used for that historical action.
For that response, the context menu exposes only Copy Message and Select All. Searching the command palette for Report AI Action in Agents Window gives no result, so I have not recovered a request ID.
What is the supported diagnostic route for locating the exact request ID and determining which integration or review component handled this operation? Is there official documentation for an approved secret-handling and export workflow in this situation? I want to retain the existing safeguards, without bypassing approval protections or posting private logs, credentials, or project details publicly.
Please advise the appropriate support channel or documented steps.
Hey @DiMcQueen-Richardson, that skip came from Cursor’s Auto-review check, whichever model you’re chatting with. It blocks agent commands that read a token and pass it into another command, unless you clearly asked for that step.
To run the export with the safeguards still on, approve it when the approval card comes up, or run it yourself in your own terminal so the agent never handles the token.
Thank you. I retrieved the Request ID from the user prompt immediately before the skipped export: 55b130d0-a5b3-4542-ae7f-59f3c164a11f. The associated response is dated 16 October 2026 at 6:04 pm AEDT (UTC+11). Please use this to confirm which review handled that turn.
I checked that turn, and it was handled entirely by Cursor’s Auto-review. The export command needed access outside the sandbox, and Auto-review had already blocked that kind of step earlier in the same chat, so it showed you an approval card with that reason. The card appeared at 6:04:17 pm and was skipped at 6:04:53 pm, so the command never ran.
If you want it to run, approve the card the next time it appears, or run the export yourself in your terminal.