Repository navigation
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info
📝 Walkthrough
Priority: ⬇️ Low Severity of issue fixed: Low Merge Risk: ⚪ Minimal · up to The change adds a shortcut that jumps to the newest agent request awaiting input. The incremental changes make the jump wait for pending workspace activation before focusing the exact tab and pane. No merge-blocking risk was identified. Pre-merge checks |
  |
| Check name | Status | Explanation | Resolution |
|---|---|---|---|
| Docstring Coverage |  |
Docstring coverage is 20.00% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff. Analyzed 20 functions across 15 files. | Write docstrings for the functions missing them to satisfy the coverage threshold. |
✅ Passed checks (4 passed)
-
Check name Status Explanation Linked Issues check Direct issue #27255has coding requirements. The PR adds the remappableworktree.jumpToLatestAttentionaction with the requestedMod+Shift+Udefault.resolveLatestAttentionThreadselects the new…Out of Scope Changes check The changes remain within issue #27255. The command registration, keybinding definition, activity-thread resolution, activation-gate refactor, documentation, unit tests, and Electron test all implemen…Title check The title clearly and concisely describes the primary change: adding a shortcut to jump to the latest agent needing input. Description check The description follows the required template and includes the user impact, implementation details, rationale, linked issue, visual proof, testing results, AI disclosure, review notes, compatibility c…
- Fix all pre-merge checks with AI
- Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.
Comment @coderabbitai help to get the list of available commands.
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
ℹ️ Review info
⚙️ Run configuration
- Configuration used: Repository UI
- Review profile: CHILL
- Plan: Advanced
- Run ID:
8e216202-6f63-443f-a325-a70b3992e187
📒 Files selected for processing (13)
docs/site/content/docs/notifications.mdxsrc/renderer/src/app-shell/app-command-handlers-latest-attention.test.tssrc/renderer/src/app-shell/app-command-handlers.tssrc/renderer/src/app-shell/latest-attention-command.tssrc/renderer/src/components/activity/activity-event-builder-context.tssrc/renderer/src/components/activity/activity-thread-actions.tssrc/renderer/src/components/activity/latest-attention-thread.test.tssrc/renderer/src/components/activity/latest-attention-thread.tssrc/shared/keybindings-latest-attention.test.tssrc/shared/keybindings/definitions-core-4.tssrc/shared/keybindings/types.tssrc/shared/plugins/plugin-command-actions.tstests/e2e/jump-to-latest-attention.spec.ts
Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 9 remain after this review.
This branch has not been deployed
1 participant
ELI5
Press Cmd+Shift+U on macOS, or Ctrl+Shift+U on Windows/Linux, to open the agent that most recently asked for input. Orca selects its workspace, tab and exact split pane. The request stays pending; the shortcut neither answers it nor grants permission.
What Changed
Before, reaching that request required finding its workspace and pane manually. After, Jump to latest agent needing input provides a configurable default shortcut that also works while a terminal has focus.
The action reads Activity's existing fresh waiting/blocked status and ranks requests by their state start time. Heartbeats and completed historical turns do not move an older request ahead. It skips deleted/archived workspaces, confirmed removed panes, closed tabs and folder paths blocked by the existing activation gate. It validates the original request's execution host before navigating, so losing a remote catalog entry cannot redirect the jump to a same-ID local workspace.
The existing Activity pane activator is extracted and reused for terminal, structured-chat, folder, floating and SSH/remote navigation. For a sleeping target it waits for the already-running restoration gate, then focuses the requested tab and leaf. A newer thread selection or a change of workspace, execution host or view cancels that deferred focus; a blocked gate or missing target never falls back to a sibling tab. With no eligible request or while a dialog owns input, the action does not consume the chord. Remapping, disabling and terminal-first policy use the existing keybinding infrastructure.
Why
This is a direct newest pending request action. Seeing a request does not resolve it, so repeated presses stay on that request until the agent resumes or a newer request arrives.
Related open PRs #21696 and #26280 respectively cycle agents needing input and drain unread results. This action deliberately has neither behavior. Reusing Activity classification and activation avoids a second status source or a separate pane-focus implementation.
Linked Issue
Fixes #27255
Visual Proof
The same isolated macOS fixture contains an older blocked request and a newer waiting request. Before: the binding is disabled and pressing the chord leaves
mainselected. After: restoring the default and pressing it selectse2e-secondaryand highlights the newer waiting agent. Only the synthetic sidebar is captured; no real workspace, account or shell content is included.The Electron test additionally asserts focus on the exact split pane's input, both pending status labels remaining visible, repeat navigation, and every test window remaining hidden.
Testing
I manually tested these changes locally — agent-driven Playwright against a built, isolated Electron app on macOS; native OS focus is not claimed.
Automated tests added/updated.
Two focused Vitest runs passed: 40 tests across selection, activation, bindings and the new command; 19 tests across the new and existing app commands. Five command tests overlap, for 54 distinct tests. Coverage includes heartbeat ordering, answered/stale/unconfirmed requests, read-but-pending requests, removed panes, closed tabs, archived/deleted workspaces, cold SSH tabs, lost remote ownership, folder path gates, floating terminals, structured chats, child visibility, malformed identities, ties, remapping and terminal-first behavior.
Review follow-up
3b76d29f: 115 tests passed across seven focused files, including nine new fixtures for deferred terminal/structured focus, stale workspace/host/view/deleted targets, blocked recovery, missing target tabs and superseding selections. Seven assertions failed against the pre-fix activator; all nine fixtures pass with the fix. Existing Activity routing, activation-gate and empty-workspace recovery tests also pass.The five new keybinding tests fail against clean upstream
5fae0fa3and pass with this change.Renderer typecheck passed:
node node_modules/typescript/bin/tsc --noEmit -p config/tsconfig.tc.web.json.Changed-code quality gate, scoped lint, formatting, commit hooks and
git diff --checkpassed.Electron main/preload/renderer build with
--mode e2epassed; the bundled CLI also built.Final keyboard acceptance passed:
SKIP_BUILD=1 ORCA_BACKGROUND_LAUNCH=1 pnpm exec playwright test tests/e2e/jump-to-latest-attention.spec.ts --config tests/playwright.config.ts --project electron-headless --workers=1.The enhanced screenshot journey initially reached its checks but timed out during fixture cleanup on a heavily loaded host. The final test fixes its renderer clock and allows 180 seconds for isolated profile setup/teardown, while preserving the normal individual UI assertion limits; it passed in 50.9 seconds total. Electron required execution outside the shell sandbox; windows stayed hidden throughout.
Full-repository lint/test/build are not claimed. The broad E2E typecheck reports the same 258 distinct diagnostics on clean upstream and this branch, with none added by these files; these include unrelated cloud dependencies and existing test fixtures. Windows/Linux native UI and live SSH were not run locally; platform keybinding and host-routing behavior have unit coverage.
The follow-up is submitted at
3b76d29f. Upstream workflow approval and CI results are reported separately from local validation; local checks are not presented as upstream CI results.AI Disclosure
Implemented and reviewed with OpenAI Codex (GPT-6), including an independent source review and agent-driven macOS Playwright acceptance. All status fixtures are synthetic; no live agent reply or permission was sent.
Review
Cross-platform modifiers use
Modand existing matching/label infrastructure. Execution hosts retain status ownership; no process, authentication, IPC, wire protocol, or provider-specific integration is added. Folder and floating workspaces remain supported. Selection runs only when the command is invoked, with existing Activity history bounds and one candidate scan; no polling or new subscriptions. UI styling is unchanged. Review findings on lost remote ownership, blocked folder paths and asynchronous recovery focus were addressed with focused fixtures.Agent skill upstream boundary
Notes
This shortcut is an app/terminal action. Forwarding it from an embedded browser guest is outside this PR. It relies on fresh reported agent status, not inferred terminal text. It does not promise to drain unread badges or resolve a request.
Source was validated on
5fae0fa3; a final upstream fetch found no overlapping changed paths. No release version or installed Orca application is changed.Author: GitHub @Mirksen. X/Twitter handle: not provided.
Checklist
pnpm lint,pnpm typecheck,pnpm testandpnpm build— scoped checks and limitations are recorded above; full CI results remain pending.