Conversation
|
Bugbot is not enabled for your account, so this pull request was not reviewed. Enable Bugbot in the Cursor dashboard to get automatic reviews on future PRs. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 1a8c337046
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| spawn: buildOmpAcpSpawnInput( | ||
| input.ompSettings, | ||
| input.cwd, | ||
| input.runtimeMode ?? "full-access", |
There was a problem hiding this comment.
Do not run text generation in OMP yolo mode
When OMP text generation calls makeOmpAcpRuntime it does not pass a runtime mode, so this fallback starts omp acp with --approval-mode yolo. The shared ACP text-generation path only collects assistant chunks and does not install a permission handler, while its prompts include user-controlled branch/diff content; if that content prompt-injects OMP into using tools, the background commit/PR/title generation flow can execute commands or modify files without any user approval. Use a restricted/approval-required mode for headless text generation instead of defaulting to full access.
Useful? React with 👍 / 👎.
| currentValue: "auto", | ||
| options: [ | ||
| { id: "off", label: "Off" }, | ||
| { id: "auto", label: "Auto", isDefault: true }, |
There was a problem hiding this comment.
Avoid dispatching fabricated OMP thinking values
For every reasoning model this fabricates off/auto and makes auto the default even though the discovered model.thinking array is the set of OMP-supported effort values. The composer dispatch path sends descriptor defaults, so an ordinary OMP turn for a model that advertises efforts like minimal/low/medium/high will try to apply thinking=auto (and users can choose off), which the ACP runtime rejects when its thinking config option only contains the advertised efforts. Build the descriptor from the discovered values, or omit the option when no real value was selected.
Useful? React with 👍 / 👎.
What Changed
Why
UI Changes
Checklist