You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Add a review mode that inspects uncommitted changes or a selected base branch and presents findings with file and line context before the user commits or pushes.
Problem to solve
T3 Code exposes diff and source-control surfaces, but review is currently an ordinary agent prompt; users lack a consistent review scope, findings list, and review completion state.
Proposed behavior
Let a user choose working-tree or base-branch scope, start a review turn, and view findings grouped by severity with links to the affected diff location. Allow marking findings addressed or dismissed and rerunning the review against the updated scope.
Acceptance criteria
Review scope is explicit before the agent runs and is shown in the resulting thread.
Findings link to a repo-relative file and line or a diff hunk when available.
The UI distinguishes unresolved, addressed, and dismissed findings and preserves them across reconnects.
Rerunning a review replaces stale findings only after the new scope is captured.
Review mode works with local and worktree projects and never silently reviews a different branch.
Affected area
Review contracts and persistence, source-control diff selection, orchestration metadata, and review panels in apps/web, apps/desktop, and apps/mobile.
Non-goals
This does not require automatic pull-request publication, merge, or a specific review model.
Alternatives considered
A free-form “review this” prompt cannot guarantee a stable diff scope or actionable finding lifecycle.
Supporting context
Official OpenAI documentation exposes /review for uncommitted or base-branch review and a review pane with findings and inline context: https://learn.chatgpt.com/docs/code-review. T3 Code contains review service code (apps/server/src/review/ReviewService.ts) and diff UI state (apps/web/src/diffPanelStore.ts), but the inspected contracts do not expose this complete user workflow.
Related upstream work: /review discussion #6695. Existing pull-request comments, verdict submission, and diff views should be retained; the requested addition is a dedicated agent review workflow with scoped findings.
Analysis basis: source inspection of T3 Code revision 9d4bb550a658. Statements about missing UI or contracts refer to that revision, not every provider-native capability or unmerged branch. This is a feature proposal based on source review, not a runtime bug reproduction.
Closing as a duplicate of #6695 (/review structured code review).
This write-up adds useful acceptance criteria (explicit scope, findings lifecycle, persistence). Please fold those details into #6695 rather than tracking the same feature in two Ideas threads.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
Uh oh!
There was an error while loading. Please reload this page.
Summary
Add a review mode that inspects uncommitted changes or a selected base branch and presents findings with file and line context before the user commits or pushes.
Problem to solve
T3 Code exposes diff and source-control surfaces, but review is currently an ordinary agent prompt; users lack a consistent review scope, findings list, and review completion state.
Proposed behavior
Let a user choose working-tree or base-branch scope, start a review turn, and view findings grouped by severity with links to the affected diff location. Allow marking findings addressed or dismissed and rerunning the review against the updated scope.
Acceptance criteria
Affected area
Review contracts and persistence, source-control diff selection, orchestration metadata, and review panels in apps/web, apps/desktop, and apps/mobile.
Non-goals
This does not require automatic pull-request publication, merge, or a specific review model.
Alternatives considered
A free-form “review this” prompt cannot guarantee a stable diff scope or actionable finding lifecycle.
Supporting context
Official OpenAI documentation exposes
/reviewfor uncommitted or base-branch review and a review pane with findings and inline context: https://learn.chatgpt.com/docs/code-review. T3 Code contains review service code (apps/server/src/review/ReviewService.ts) and diff UI state (apps/web/src/diffPanelStore.ts), but the inspected contracts do not expose this complete user workflow.Related upstream work: /review discussion #6695. Existing pull-request comments, verdict submission, and diff views should be retained; the requested addition is a dedicated agent review workflow with scoped findings.
Analysis basis: source inspection of T3 Code revision
9d4bb550a658. Statements about missing UI or contracts refer to that revision, not every provider-native capability or unmerged branch. This is a feature proposal based on source review, not a runtime bug reproduction.All reactions