Skip to content

Launch an agent with a first prompt from its setup panel - #36659

Merged
brennanb2025 merged 24 commits into
mainfrom
brennanb2025/launch-prompt
Nov 2, 2026
Merged

brennanb2025 merged 24 commits into
mainfrom
brennanb2025/launch-prompt

Conversation

@brennanb2025

@brennanb2025 brennanb2025 commented Oct 31, 2026 •

Copy link
Copy Markdown
Contributor
Files Added Deleted Net
Test 37 $\color{#1a7f37}{\Huge{\mathbf{+}}}$​4457 $\color{#cf222e}{\Huge{\mathbf{−}}}$​26 $\color{#1a7f37}{\Huge{\mathbf{+}}}$​4431
Prod 73 $\color{#1a7f37}{\Huge{\mathbf{+}}}$​1715 $\color{#cf222e}{\Huge{\mathbf{−}}}$​441 $\color{#1a7f37}{\Huge{\mathbf{+}}}$​1274

Problem

Starting an agent from the tab bar's + menu always opens it empty: to give it a task you have to wait for the chat or terminal to open and then type. The agent's › setup panel (model, effort, permissions, arguments) had no place to say what the agent should do.

What changes for users

The › setup panel now has a one-line prompt box above its Launch buttons, with the placeholder "Add a prompt (optional): / for skills, @ for files" (translated in every app language).

  • It is the same composer as a chat: / lists the skills and commands of the workspace the agent will run in, @ finds files there, and Attach adds files. The microphone uses the same dictation button as chat, including toggle and hold behavior and the same availability rules; dictated text enters the prompt like typing. The / and @ lists open above the setup panel. Attach and the microphone show while the box has focus or text, and remain visible during active dictation. The English empty box stays one line; the known long-placeholder overflow is listed below.
  • It grows as you type to about six lines, then scrolls.
  • Enter launches the panel's default way (the highlighted button) with the prompt; Shift+Enter adds a line. Either Launch button sends the prompt its own way. With nothing typed, both buttons launch exactly as before.
  • Clicking an agent's row (not ›) still launches at once with no prompt.
  • Escape in the composer closes its open / or @ list. If focus moves away while a list remains open, Escape can leave the setup panel open; that known edge case is deferred below.
  • The typed prompt and the options you pick stay when the panel closes and reopens, separately for each workspace (and host) and agent. Another workspace starts with its own. They last for the current app window session and are not restored after a restart.
  • A search result in the + menu (for example "opus high") keeps its own options, as before; opening a setup without changing anything saves nothing.
  • If a workspace is renamed or moves to another host, its typed prompt and options follow it.
  • If a launch fails after you sent the prompt, the text comes back to the setup panel, unless the new chat already holds it, so it never appears twice. If the workspace or project was removed in the meantime, it isn't brought back.

How

  • The panel renders the existing NativeChatComposer with a launch transport whose send is the launch, the same pattern the pending-worktree chat uses.
  • The composer gets one opt-in prop, launchPrompt, that turns on:
    • the compact layout, on the app's popover surface instead of the chat canvas;
    • Attach and dictation as its controls;
    • no /resume;
    • skill lookup from the launch workspace's folder, since there is no tab yet.
  • The prompt is an ordinary rich composer draft kept in memory only, keyed by workspace, host and agent. The setup panel's options live beside it, one slot per search preset. No setup record is retained until an authored prompt, option change or pending attachment needs it; merely showing a row creates no record or closed-row subscription.
  • When a workspace's identity changes (rename, a folder pinned to a host, an SSH target re-adopted), the existing catalog and rename events move that workspace's setup to its new key.
  • The handle gains submit() so the Launch buttons send through the composer's own send path. That path also resolves pastes and pending files. The launch reports whether the new chat took the text, so a failed launch returns the prompt only when nothing else holds it.
  • The prompt rides the launch's existing prompt field (useTabAgentMenuLaunch → launchAgentInNewTab), so terminal and chat launches deliver it as they already do. There is no new wire field or saved schema. Setup drafts skip ordinary draft persistence; existing attachment and long-paste preparation can still create their normal files. Unshipped saved setup residue is ignored rather than automatically deleted; it can still occupy shared journal space and bulk-load bytes.

Differences from the common pattern

  • Images go by file path (intended). In the common pattern, a new thread's first message can carry images uploaded before the session exists. Orca's launch prompt is text on both surfaces, so the box sets acceptsImages: false, and an attached image or file goes in as its path, as it already does for agents without image input. Uploading images ahead of a launch would be a new capability.
  • Attach is off for paired-server workspaces (intended). There is no session yet to upload a file into, so Attach is disabled for a workspace on a paired server. Typing, / and @ still work.
  • The prompt and options last for the app session only (intended). That matches how main already kept setup options. Keeping them across a restart was ruled optional.
  • Search presets keep their own options (intended). The common pattern has no search presets; this keeps main's per-preset behavior, scoped to the workspace.
  • A setup that was typed in keeps its record after the box is cleared (intended). Sending clears the box while the launch is still in flight, and a failed launch returns the text to that record. The common pattern drops an emptied draft because it has no in-flight launch to return to.

Tests

  • AgentLaunchSetupPanel.prompt.test.tsx runs the real composer: Enter launches the default way with the prompt; the other button launches its own way; an empty box launches without a prompt; the box clears after a launch; the draft never reaches draft storage; the German placeholder.
  • AgentLaunchSetupButton.retention.test.tsx: close and reopen keep the prompt and options per workspace and agent; search presets keep their own options and opening saves nothing; rendering 100 or 1,000 unopened rows creates nothing, and catalog and rename updates visit only the setups that were typed in.
  • NativeChatComposerActions.test.tsx: the launch prompt has Attach and dictation, no duplicate Send button, and the microphone invokes the existing dictation toggle.
  • NativeChatAutocompleteMenus.test.tsx: the / and @ lists sit above a popover and are unchanged in an ordinary chat.
  • use-native-chat-skills.test.ts covers the no-tab skill lookup from the launch workspace's folder.
  • Store tests cover the rename, folder-catalog, SSH re-adoption and removal paths, and the failed-launch return.

Verification and known follow-ups

CI on 2bd0c872e5134f6ed2610edd10aa2f7002a43835 is green: 35 checks passed and 23 were skipped, including all 12 unit shards, static analysis/typecheck, Mobile verification, macOS/Windows packaging, changed E2E specs, managed SSH and all three WSL distro checks. The inherited Mobile Hermes and earlier main-test blockers are retired. Renderer/store tests cover launch inputs and refusal, rich failed-send recovery, workspace/host/agent partitioning, per-preset options, positive rename/host association and committed-removal handling. The real-trigger retention contract uses 100 and 1,000 unopened rows and visits only five authored sources on each folder-catalog apply.

The main-fix merge 4dd740fc19b6facbb1b6e62d73d3a72d011fb898 brings in #37370 and the other intervening main updates. The existing merge-check reviewer checked both conflict resolutions against both parents and affected callers and found no merge-resolution defect. Main’s agent-reference serialization is preserved beside this PR’s capture/consume/recovery, using equivalent synchronous reads of the same draft. The test retains main’s production handle and this PR’s edit-session ref; the auto-merged goal, paste and catalog paths remain intact. The owner reports six named test files passed (129/129) through tc-gate and the merge hook passed. The dictation setup screenshot remains at 5e26f380f79684146e52a85d78ece17823b721d3; a main merge preserves that evidence. No new app run was performed. The subsequent test-only commit 2bd0c872e5134f6ed2610edd10aa2f7002a43835 adds the missing agent-reference method to one test fixture, matching main’s expanded handle interface. It changes no assertions or shipped code. The same reviewer accepted the CI repair; the named test passed 3/3 and the standard hook passed. This non-shipped change also preserves QA and reference evidence.

The conflict-blocking main merge 8b2bea29cbc390fef3be4a8c9fcc4a44c512581a was narrowly checked against both parents and affected callers, with no merge-resolution defect. Its targeted tc-gate run reports 26 files / 309 tests passing. Main’s own custom-editor lint failure then required a further clean main merge, f1509e5f89cd0265b291a222286a3970e0a30953, bringing in the existing fix from #37314. Its delta matches main’s seven commits exactly; only unrelated English translation keys overlap this PR. The earlier CLEAN merge check carries. The subsequent test-only commit a013c9d31622fa626f8d47eddccf60af14fc9fc9 updates three promise annotations to match main’s send result; the continued reviewer confirmed that no assertions or shipped behavior changed. The owner reports 27/27 targeted tests passing. The existing main plugin-test failures then required the clean main merge fc93c63a554e1a133a2c2dbb68958b743eb7be30, bringing in the main fixes from #37324 and #37326. Its delta is exactly main’s nine intervening commits; the only overlap with this PR is 11 locale JSONs. All locales parse, main’s semantic changes are preserved, and every PR-added locale change remains intact. No incoming change touches the launch/composer/draft code paths, so the prior CLEAN narrow merge review and reference/UI evidence carry. The remaining inherited Electron probe timeout required the clean main merge faa7d0d18fc8626e4ad5b99e2282e7c447d0fbb5, bringing in #37329. Its delta equals main’s two intervening commits across 88 paths with zero overlap with PR files. Every incoming tree entry matches main and every other entry matches the PR parent; none touches the launch/composer code paths. The prior CLEAN narrow review and reference/UI evidence carry. Existing branch UI evidence carries across these merges; no additional app run was performed. Main’s new /goal chip now also reaches the setup composer; that interaction was source-checked, without a dedicated rendered test.

Brennan requested the existing microphone in the setup prompt after the earlier readiness check. The narrow addition at 5e26f380f79684146e52a85d78ece17823b721d3 reuses its existing handlers and dictation hook, with the action row kept visible while dictation is active. The existing reviewer checked the addition and affected callers and found no defect: ordinary composer behavior and the shared toggle/hold/cancel/disabled paths are unchanged, and text insertion still reaches the launch prompt. The owner reports the standard hook and three named test files passed (27/27). A non-blocking coverage concern remains: no explicit test asserts that the row stays visible during active dictation after focus leaves. The one whole-window screenshot at the dictation commit shows the focused empty setup prompt with Attach left and the microphone right, above both Launch buttons. It was captured through qa-drive in an isolated background copy and inspected. The mic is visible but disabled because dictation is not configured in that fresh profile; the existing availability rules are preserved. No audio or live-dictation run is claimed.

The native-chat UI QA table calls for renderer tests plus branch screenshots. macOS screenshots cover the empty and focused box, growth/scrolling, retention, search presets, light theme, and clickable / and @ lists above the setup panel. These were not live agent-delivery runs: Claude not exercised in a running chat/terminal launch; Codex not exercised in a running chat/terminal launch. Automated launch fixtures and CI cover those shared inputs and outcomes. No restart-retention claim is made.

  • Known follow-up — translated empty hint: long placeholders overflow sideways. On Windows at pre-merge 06d27575c2180450bc7265c8e1991955207f3b2f, French shows a horizontal scrollbar and clips the tail of the hint; German overruns the padding. Typing removes the hint and both Launch buttons stay usable. The translated-placeholder QA step failed and is deferred, tracked in STA-13245.
  • Known follow-up — Escape after focus leaves an open list: an expanded / or @ list can keep the setup panel open when Escape is pressed outside the composer. Deferred as low priority, tracked in STA-13404.
  • Spanish uses “skills” where its / list says “Habilidades”, tracked in STA-13230.
  • A search row can still say “High” after its retained setup was changed to Low; this is existing per-preset label behavior.
  • Temporary inherited file-lookup limitation: the existing same-ID workspace/host-twin @ lookup issue is tracked in STA-13130. The fresh-profile model-catalog issue is tracked separately in STA-13228. Neither is fixed here. Existing server-side folder host-reassignment coverage also lacks its own folder test; this is a coverage concern, not a demonstrated PR regression.

Current requested dictation screenshot (whole window, 5e26f380):

Launch prompt with existing dictation button

Screenshots, evidence carry and the deferred layout failure are detailed in the review status comment.

The agent setup panel in the new-tab menu gets an optional prompt box above
its Launch buttons. It is the chat composer in a compact launch layout: one
line that grows to six, Attach as its only control, / listing the launch
workspace's skills and commands and @ finding its files. Enter launches the
panel's default way, either Launch button sends the prompt its own way, and
with nothing typed both launch exactly as before. The prompt rides the
existing launch prompt field and is kept in memory only until the launch.
@brennanb2025
brennanb2025 marked this pull request as ready for review October 31, 2026 09:46
@coderabbitai

coderabbitai Bot commented Oct 31, 2026 •

Copy link
Copy Markdown
Contributor

Review in Change Stack →

Note

Reviews paused

It looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the reviews.auto_review.auto_pause_after_reviewed_commits setting.

Use the following commands to manage reviews:

  • @coderabbitai resume to resume automatic reviews.
  • @coderabbitai review to trigger a single review.

Use the checkboxes below for quick actions:

  • Resume reviews
  • Trigger review
Recent review info
Run configuration
  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: Repository UI
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 2acd17db-af7e-4694-8986-25acdec6ed4b
Commits

Reviewing files that changed from the base of the PR and between 03c0d13 and 31c4c1b.

Files selected for processing: 41
  • src/renderer/src/components/agent/AgentLaunchPickerMenu.test.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupButton.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupPanel.prompt.test.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupPanel.tsx
  • src/renderer/src/components/native-chat/inline-composer-scaling.test.tsx
  • src/renderer/src/components/native-chat/native-chat-composer-destination.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-image-check.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-load.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-memory.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-removal.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-store.ts
  • src/renderer/src/components/native-chat/native-chat-composer-pasted-text-references.ts
  • src/renderer/src/components/native-chat/native-chat-draft-cache.test.ts
  • src/renderer/src/components/native-chat/native-chat-launch-draft-continuity.ts
  • src/renderer/src/components/native-chat/native-chat-launch-draft-source.ts
  • src/renderer/src/components/native-chat/native-chat-launch-draft-transition.ts
  • src/renderer/src/components/native-chat/native-chat-paste-destination.ts
  • src/renderer/src/components/native-chat/native-chat-pending-attachment-cache.ts
  • src/renderer/src/components/native-chat/use-native-chat-composer-attachments.ts
  • src/renderer/src/components/native-chat/use-native-chat-composer-submit.test.tsx
  • src/renderer/src/components/native-chat/use-native-chat-draft.test.tsx
  • src/renderer/src/components/sidebar/HostRemoveDialog.test.tsx
  • src/renderer/src/hooks/ipc-events/worktree-event-runtime-activation.test.ts
  • src/renderer/src/hooks/ipc-events/worktree-event-runtime.ts
  • src/renderer/src/store/folder-workspaces/folder-workspace-catalog-actions.ts
  • src/renderer/src/store/folder-workspaces/folder-workspace-mutations.ts
  • src/renderer/src/store/projects/project-catalog-null-field-ingest.test.ts
  • src/renderer/src/store/repos/all-host-repo-catalog-actions.ts
  • src/renderer/src/store/repos/repo-catalog-actions.ts
  • src/renderer/src/store/repos/repo-catalog-merge.ts
  • src/renderer/src/store/repos/repo-removal.test.ts
  • src/renderer/src/store/repos/repo-removal.ts
  • src/renderer/src/store/repos/runtime-repo-catalog-actions.ts
  • src/renderer/src/store/slices/repos-catalog-reconciliation.test.ts
  • src/renderer/src/store/slices/runtime-status-catalog-identity.test.ts
  • src/renderer/src/store/slices/runtime-status.ts
  • src/renderer/src/store/slices/superseded-ssh-repo-rows.ts
  • src/renderer/src/store/slices/workspace-cleanup-wrong-host-removal-guard.test.ts
  • src/renderer/src/store/slices/worktrees/listing/stale-listing-purges-created-worktree.test.ts
  • src/renderer/src/store/slices/worktrees/teardown/removed-worktree-chat-drafts.ts
  • src/renderer/src/store/slices/worktrees/teardown/worktree-state-before-removal.ts
Files with no reviewable changes: 1
  • src/renderer/src/store/slices/runtime-status.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 8 remain after this review.

Walkthrough

The change adds workspace-scoped agent launch prompts with skill and file references, launch-specific composer behavior, and controlled prompt submission. Launch callbacks now report acceptance so menus and setup popovers can retain refused launches. Native-chat draft state tracks submissions, recovery, attachments, and scope transitions. Workspace, host, project, folder, runtime, and worktree changes migrate or retire matching draft state. Tests cover launch, recovery, reassignment, and teardown paths.

Merge Risk: Low · up to 31c4c

The launch-prompt feature mostly works as intended. Three minor edge cases remain: drafts may not follow an SSH host whose target ID needs percent-encoding, the setup popover's Escape handling can fail to close it, and a caller-supplied prompt can cause the typed setup prompt to be dropped. Fixing them before merge is advisable, but none blocks the change.

Pre-merge checks: Passed 3 · Failed 2

Failed Failed checks: 2 warnings

Check name Status Explanation Resolution
Docstring Coverage Warning Docstring coverage is 28.09% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documenta… Write docstrings for the functions missing them to satisfy the coverage threshold.
Description check Warning The description follows the template's main sections (ELI5 is absent, but Problem, What changes for users, How, Tests, Risk, Verification and known follow-ups are present). It explains the change and … Add an ELI5 section; link an actual issue under Linked Issue (replace 'Fixes #'); provide a before/after visual for the UI change in Visual Proof (the current single whole-window screenshot does not show a before state); state the Risk leve…
Passed Passed checks: 3 passed
Check name Status Explanation
Title check Passed The title "Launch an agent with a first prompt from its setup panel" matches the main change: the agent setup panel gains a prompt box whose text is sent with the launch. The changeset also covers sup…
Linked Issues check Passed Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check Passed Check skipped because no linked issues were found for this pull request.
Full details: Title check

Explanation

The title "Launch an agent with a first prompt from its setup panel" matches the main change: the agent setup panel gains a prompt box whose text is sent with the launch. The changeset also covers supporting work (draft scopes, skill lookup, launch-result handling), but the title names the primary user-facing change and is specific enough for a teammate scanning history.

Full details: Docstring Coverage

Explanation

Docstring coverage is 28.09% which is insufficient. The required threshold is 80.00%. Docstring coverage is scoped to functions touched by this diff with cyclomatic complexity of at least 2. Documentation-only edits and deletions within surviving functions are included; unchanged diff context and fully deleted functions are excluded. Analyzed 89 functions across 86 files.

Resolution

Write docstrings for the functions missing them to satisfy the coverage threshold.

Full details: Description check

Explanation

The description follows the template's main sections (ELI5 is absent, but Problem, What changes for users, How, Tests, Risk, Verification and known follow-ups are present). It explains the change and mechanism in depth, lists tests, and documents known follow-ups. However, the template's required sections are not clearly matched: there is no ELI5 section, no 'Linked Issue' with a 'Fixes #' reference (only a PR objective reference, no issue link), no explicit 'Risk' level (Level:) or backout plan, no 'AI Disclosure' content, no 'Review' or 'Agent skill upstream boundary' checkbox, and the Visual Proof is present (screenshot) but the before/after pair required by the template is not provided. Testing checkboxes are not marked. The description is mostly complete but several required template items are missing, so it is not fully aligned with the template.

Resolution

Add an ELI5 section; link an actual issue under Linked Issue (replace 'Fixes #'); provide a before/after visual for the UI change in Visual Proof (the current single whole-window screenshot does not show a before state); state the Risk level (Low/Medium/High) and a back-out plan; complete the Testing checkboxes (manual testing and automated tests) and state which platforms were tested; complete or mark N/A the AI Disclosure section for non-team contributors; and complete or mark N/A the Agent skill upstream boundary checkbox.

Finishing Touches
Generate docstrings
  • Commit to this branch
  • Create a new PR
  • 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.

Share

@coderabbitai coderabbitai Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actionable comments posted: 1


Review info
Run configuration
  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: Repository UI
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 86c111ad-bf23-4d91-ae68-0aee05f9eeaf
Commits

Reviewing files that changed from the base of the PR and between fdfcfb3 and bf53f04.

Files selected for processing: 24
  • src/renderer/src/components/agent/AgentLaunchPromptComposer.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupButton.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupFooter.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupPanel.prompt.test.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupPanel.tsx
  • src/renderer/src/components/agent/use-agent-launch-picker-inventory.ts
  • src/renderer/src/components/native-chat/NativeChatComposerActions.tsx
  • src/renderer/src/components/native-chat/NativeChatComposerField.tsx
  • src/renderer/src/components/native-chat/NativeChatComposerPane.tsx
  • src/renderer/src/components/native-chat/native-chat-composer-draft-memory.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-store.ts
  • src/renderer/src/components/native-chat/native-chat-composer-types.ts
  • src/renderer/src/components/native-chat/native-chat-skill-discovery-context.ts
  • src/renderer/src/components/native-chat/queued-edit-trigger-scaling.test.tsx
  • src/renderer/src/components/native-chat/use-native-chat-composer-command-picker.ts
  • src/renderer/src/components/native-chat/use-native-chat-composer-edit-session.ts
  • src/renderer/src/components/native-chat/use-native-chat-composer-keydown.ts
  • src/renderer/src/components/native-chat/use-native-chat-picker-state.ts
  • src/renderer/src/components/native-chat/use-native-chat-resume-command.ts
  • src/renderer/src/components/native-chat/use-native-chat-skills.test.ts
  • src/renderer/src/components/native-chat/use-native-chat-skills.ts
  • src/renderer/src/components/tab-bar/QuickLaunchButton.tsx
  • src/renderer/src/i18n/locales/en.json
  • src/shared/agent-launch-picker.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 3 remain after this review.

@coderabbitai coderabbitai Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actionable comments posted: 1


Review info
Run configuration
  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: Repository UI
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 56ab62fe-af74-4717-a8e5-f6c0b374fd8a
Commits

Reviewing files that changed from the base of the PR and between 054f7cc and e859497.

Files selected for processing: 30
  • src/renderer/src/components/agent/AgentLaunchPickerMenu.tsx
  • src/renderer/src/components/agent/AgentLaunchPromptComposer.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupButton.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupFooter.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupPanel.tsx
  • src/renderer/src/components/native-chat/native-chat-composer-draft-comparison.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-memory.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-store.ts
  • src/renderer/src/components/native-chat/native-chat-composer-types.ts
  • src/renderer/src/components/native-chat/native-chat-structured-composer-dispatch.ts
  • src/renderer/src/components/native-chat/use-native-chat-composer-submit.test.tsx
  • src/renderer/src/components/native-chat/use-native-chat-structured-composer-send.test.tsx
  • src/renderer/src/components/native-chat/use-native-chat-structured-composer-send.ts
  • src/renderer/src/components/tab-bar/QuickLaunchButton.tsx
  • src/renderer/src/components/tab-bar/TabBarCreateEntry.tsx
  • src/renderer/src/components/tab-bar/TabBarCreateEntryRow.tsx
  • src/renderer/src/components/tab-bar/tab-bar-create-menu-contract.ts
  • src/renderer/src/components/tab-bar/tab-create-entry-props.ts
  • src/renderer/src/components/tab-bar/use-tab-bar-create-menu-controller.ts
  • src/renderer/src/lib/launch-agent-in-new-tab-host-agents.ts
  • src/renderer/src/lib/launch-agent-picker-prompt.test.tsx
  • src/renderer/src/lib/launch-agent-structured-chat-test-fixture.ts
  • src/renderer/src/lib/new-agent-prompt-outcome.ts
  • src/renderer/src/lib/structured-agent-session-launch-prompt.test.ts
  • src/renderer/src/lib/structured-agent-session-launch-prompt.ts
  • src/renderer/src/store/folder-workspaces/folder-workspace-mutations.ts
  • src/renderer/src/store/repos/repo-removal.test.ts
  • src/renderer/src/store/slices/folder-workspace-owner-routed-mutations.test.ts
  • src/renderer/src/store/slices/worktrees/teardown/remove-worktree-shared-records.test.ts
  • src/renderer/src/store/slices/worktrees/teardown/removed-worktree-chat-drafts.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 4 remain after this review.

@coderabbitai coderabbitai Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Caution

Some comments are outside the diff and can’t be posted inline due to GitHub limitations.

⚠️ Outside diff range comments (2)

Minor · Await the launch result before dismissing the picker menu. · AgentLaunchPickerMenu.tsx:171-179

src/renderer/src/components/agent/AgentLaunchPickerMenu.tsx:171-179
Minor · Functional Correctness · Quick win

Await the launch result before dismissing the picker menu.

result !== false accepts a pending promise immediately. If prompt delivery later resolves false, item.current?.click() has already dismissed the Radix menu. Await runLaunch before setting dismissing.current and clicking the row. This fixes the picker dismissal only; the setup-popover close is a separate issue.

Suggested fix
-      onUse={(draft) => {
-        const result = runLaunch(pickerLaunchRow(entry, draft.surface), draft, entry)
+      onUse={async (draft) => {
+        const result = await runLaunch(pickerLaunchRow(entry, draft.surface), draft, entry)
         if (result !== false) {
Minor · Keep the setup popover open until a pending launch result… · AgentLaunchSetupButton.tsx:164-176

src/renderer/src/components/agent/AgentLaunchSetupButton.tsx:164-176
Minor · Functional Correctness · Quick win

Keep the setup popover open until a pending launch result settles.

AgentLaunchSetupButton closes its popover when onUse returns a pending Promise, because accepted !== false is true for a Promise. For a nonempty setup prompt, the launch path returns a prompt-ownership promise. If that promise resolves false, the popover has already closed. The user gets no indication in this path that the prompt was refused. The draft stays in sessionDrafts, so the text is not lost. Only the refusal state is hidden.

The button should wait for a pending result before it decides whether to close.

Suggested fix
           onUse={(draft) => {
             const accepted = onUse(draft)
+            if (accepted instanceof Promise) {
+              return accepted.then((ok) => {
+                if (ok !== false) {
+                  changeOpen(false)
+                }
+                return ok
+              })
+            }
             if (accepted !== false) {
               changeOpen(false)
             }
             return accepted
           }}

Review info
Run configuration
  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: Repository UI
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 8f59b3c3-87e2-4e1e-b270-b3594fcc080f
Commits

Reviewing files that changed from the base of the PR and between e859497 and 03c0d13.

Files selected for processing: 25
  • src/renderer/src/components/agent/AgentLaunchSetupPanel.prompt.test.tsx
  • src/renderer/src/components/native-chat/native-chat-composer-draft-removal.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-store.ts
  • src/renderer/src/components/native-chat/native-chat-pending-attachment-cache.ts
  • src/renderer/src/hooks/ipc-events/worktree-event-runtime-activation.test.ts
  • src/renderer/src/hooks/ipc-events/worktree-event-runtime.ts
  • src/renderer/src/lib/launch-agent-picker-prompt.test.tsx
  • src/renderer/src/store/project-groups/project-group-mutations.ts
  • src/renderer/src/store/project-groups/project-group-removal-state.ts
  • src/renderer/src/store/projects/project-catalog-null-field-ingest.test.ts
  • src/renderer/src/store/projects/project-host-setup-actions.ts
  • src/renderer/src/store/repos/repo-removal.test.ts
  • src/renderer/src/store/repos/repo-removal.ts
  • src/renderer/src/store/slices/project-group-removal-targets.test.ts
  • src/renderer/src/store/slices/runtime-status-catalog-identity.test.ts
  • src/renderer/src/store/slices/runtime-status-snapshot.ts
  • src/renderer/src/store/slices/runtime-status.ts
  • src/renderer/src/store/slices/workspace-cleanup-wrong-host-removal-guard.test.ts
  • src/renderer/src/store/slices/worktrees-background-removal.test.ts
  • src/renderer/src/store/slices/worktrees/listing/authoritative-worktree-removal-memory.ts
  • src/renderer/src/store/slices/worktrees/listing/fetched-worktree-merge.ts
  • src/renderer/src/store/slices/worktrees/listing/stale-listing-purges-created-worktree.test.ts
  • src/renderer/src/store/slices/worktrees/teardown/remove-worktree.ts
  • src/renderer/src/store/slices/worktrees/teardown/removed-worktree-chat-drafts.ts
  • src/renderer/src/store/slices/worktrees/teardown/worktree-state-before-removal.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 4 remain after this review.

@coderabbitai coderabbitai Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Actionable comments posted: 1


Review info
Run configuration
  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: Repository UI
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 2acd17db-af7e-4694-8986-25acdec6ed4b
Commits

Reviewing files that changed from the base of the PR and between 03c0d13 and 31c4c1b.

Files selected for processing: 41
  • src/renderer/src/components/agent/AgentLaunchPickerMenu.test.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupButton.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupPanel.prompt.test.tsx
  • src/renderer/src/components/agent/AgentLaunchSetupPanel.tsx
  • src/renderer/src/components/native-chat/inline-composer-scaling.test.tsx
  • src/renderer/src/components/native-chat/native-chat-composer-destination.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-image-check.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-load.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-memory.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-removal.ts
  • src/renderer/src/components/native-chat/native-chat-composer-draft-store.ts
  • src/renderer/src/components/native-chat/native-chat-composer-pasted-text-references.ts
  • src/renderer/src/components/native-chat/native-chat-draft-cache.test.ts
  • src/renderer/src/components/native-chat/native-chat-launch-draft-continuity.ts
  • src/renderer/src/components/native-chat/native-chat-launch-draft-source.ts
  • src/renderer/src/components/native-chat/native-chat-launch-draft-transition.ts
  • src/renderer/src/components/native-chat/native-chat-paste-destination.ts
  • src/renderer/src/components/native-chat/native-chat-pending-attachment-cache.ts
  • src/renderer/src/components/native-chat/use-native-chat-composer-attachments.ts
  • src/renderer/src/components/native-chat/use-native-chat-composer-submit.test.tsx
  • src/renderer/src/components/native-chat/use-native-chat-draft.test.tsx
  • src/renderer/src/components/sidebar/HostRemoveDialog.test.tsx
  • src/renderer/src/hooks/ipc-events/worktree-event-runtime-activation.test.ts
  • src/renderer/src/hooks/ipc-events/worktree-event-runtime.ts
  • src/renderer/src/store/folder-workspaces/folder-workspace-catalog-actions.ts
  • src/renderer/src/store/folder-workspaces/folder-workspace-mutations.ts
  • src/renderer/src/store/projects/project-catalog-null-field-ingest.test.ts
  • src/renderer/src/store/repos/all-host-repo-catalog-actions.ts
  • src/renderer/src/store/repos/repo-catalog-actions.ts
  • src/renderer/src/store/repos/repo-catalog-merge.ts
  • src/renderer/src/store/repos/repo-removal.test.ts
  • src/renderer/src/store/repos/repo-removal.ts
  • src/renderer/src/store/repos/runtime-repo-catalog-actions.ts
  • src/renderer/src/store/slices/repos-catalog-reconciliation.test.ts
  • src/renderer/src/store/slices/runtime-status-catalog-identity.test.ts
  • src/renderer/src/store/slices/runtime-status.ts
  • src/renderer/src/store/slices/superseded-ssh-repo-rows.ts
  • src/renderer/src/store/slices/workspace-cleanup-wrong-host-removal-guard.test.ts
  • src/renderer/src/store/slices/worktrees/listing/stale-listing-purges-created-worktree.test.ts
  • src/renderer/src/store/slices/worktrees/teardown/removed-worktree-chat-drafts.ts
  • src/renderer/src/store/slices/worktrees/teardown/worktree-state-before-removal.ts
Files with no reviewable changes: 1
  • src/renderer/src/store/slices/runtime-status.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 8 remain after this review.

Rendering an agent row in the launch menu or search no longer creates a
setup record. The row derives its scope from its workspace, host and
agent without touching session state; the record is created by the first
typed prompt, option change or pending attachment, matching main, which
kept only authored setups. Catalog applies, readoptions and renames now
visit authored setups only.

Option listeners use the existing per-scope draft listeners and subscribe
only while the setup popover is open. Also return the runtime-status
files to main, which only carried leftovers from a removed hook.
Deleting a project group no longer retires in-flight setup submissions
for its folders: folder keys are never derived again, so it changed
nothing users could see. The group mutation, its removal state and the
prompt-delivery type widening go back to main's content, with the test
that exercised the revocation.

Test fixtures catch up with the setup prompt composer: the TabBar window
stubs gain event listeners, the picker inventory's mock store gains agent
status, create-rejection handbacks expect inComposer, and the folder
removal test expects a same-ID survivor's chip to stay, as on main.
Opening a search result's setup no longer writes its preset over the
workspace's setup. As on main, each preset keeps its own options slot,
now inside the workspace/host/agent setup, keyed by what the search
parsed rather than by defaults; the plain row keeps its own slot. Only a
real change saves, into that entry's slot, and creates the setup then.
The typed prompt stays one per workspace, host and agent.
Main moved the native chat agent profiles module under shared/native-chat,
so the setup prompt composer imports it from there. Files that imported
one module twice now import it once; nothing else changes.
The catalog continuity test no longer imports the main process's SSH
target reassignment, which pulled the whole main graph into the web
program; it builds the reassigned folder inline. Test fixtures and spies
now match current types, the edit-session tests pass a composer ref as
production does, and the setup panel again accepts callers that return
no launch answer. The pending-chip owner filter drops an unused name.
No shipped behavior changes.
In the agent setup popover, the / and @ lists opened underneath the
panel, so they could not be seen or clicked. A list opened inside a
popover now takes the menu layer that selects and dropdowns already use
above popovers; an ordinary chat's lists keep their container and layer.

The prompt placeholder is translated in every locale, and folder draft
continuity takes the fetched catalog itself instead of reading its
folder list field.
With a longer translation, the empty setup prompt's placeholder wrapped
to a second line that the one-line box cut off. The empty launch prompt
now lets its placeholder's wrapped height count, so the box fits it;
English still fits one line, and an ordinary chat's box is unchanged.
@brennanb2025

brennanb2025 commented Nov 1, 2026 •

Copy link
Copy Markdown
Contributor Author

Starting an agent from the + menu previously opened it empty. This PR adds the existing chat composer to the agent's › setup panel: one line that grows to six, with workspace skills, file mentions, Attach, dictation and the existing chat/terminal Launch buttons. Enter uses the default Launch button; direct row launch and empty launch keep their previous behavior.

The prompt is retained through closing/reopening per workspace, host and agent for the current window session. Search presets retain separate option slots; opening a preset saves nothing. Accepted workspace renames/host moves carry the draft forward. A failed launch returns the complete draft only when the new chat does not already own it. Setup drafts use the ordinary rich composer store without restart persistence; existing file/paste preparation is unchanged. Images are passed by file path and Attach remains disabled before a paired-server session exists (intended differences). Unshipped setup residue is ignored rather than deleted and may still occupy journal space/load bytes.

Brennan-requested dictation addition

Brennan asked on 11-01: “For 36659 why doesn’t it show the dictation option? it should.” Commit 5e26f380f79684146e52a85d78ece17823b721d3 shows the existing microphone beside Attach, on the right as in chat. It retains the same toggle/hold handlers and dictationDisabled rules. The action row appears on focus or content and stays visible during active dictation so Stop cannot disappear after blur. The existing hook targets this editor, so dictated text enters the prompt and the existing Enter/Launch path sends it. The existing reviewer checked the addition and affected callers and found no defect: ordinary composer behavior and the shared toggle/hold/cancel/disabled paths are unchanged, and text insertion still reaches the launch prompt. The owner reports the standard hook and three named test files passed (27/27). A non-blocking coverage concern remains: no explicit test asserts that the row stays visible during active dictation after focus leaves.

Brennan then narrowed the running-app evidence: “For the dictation one, just do a quick screenshot, like one screenshot”. The one whole-window screenshot at the dictation commit shows the focused empty setup prompt with Attach left and the microphone right, above both Launch buttons. It was captured through qa-drive in an isolated background copy and inspected. The mic is visible but disabled because dictation is not configured in that fresh profile; the existing availability rules are preserved. No audio or live-dictation run is claimed. No microphone/audio or agent launch is claimed. This explicitly requested product addition is a narrow post-readiness change, not an additional review round; six historical rounds remain six.

Review and verification

  • Full-path review, six historical rounds. The final three-seat panel settled the design outcome READY at 06d27575c2180450bc7265c8e1991955207f3b2f, with the P2/P3 follow-ups below; this is not a claim that every visual check passed. The additional finishing loop was not a permitted extra review round and does not reset the count.
  • Fixed findings include lost prompts on refusal, launch-path/Enter parity, exact selected skill paths, independent failed-send recovery, workspace/host continuity, allocation for unopened rows, preset options being overwritten on open, and autocomplete lists hidden under the setup popover.
  • CI on 2bd0c872e5134f6ed2610edd10aa2f7002a43835 is green: 35 checks passed and 23 were skipped, including all 12 unit shards, static analysis/typecheck, Mobile verification, macOS/Windows packaging, changed E2E specs, managed SSH and all three WSL distro checks. The inherited Mobile Hermes and earlier main-test blockers are retired. Renderer/store/launch fixtures cover shared inputs and outcomes, including Claude/Codex launch cases; the real-trigger N=100/1,000 retention contract leaves unopened rows unallocated and visits exactly five authored setups per folder apply.
  • UI classification is qa=tests: renderer tests plus branch-only screenshot runs. No agent was launched during those runs. Claude not exercised in a running chat/terminal launch; Codex not exercised in a running chat/terminal launch. No restart retention was tested or promised.
  • Final whole-mechanism reference comparison at b94be72f9960e15dd45063fe66d1f33a0154f2d9 supports refs=done; its broken import was fixed by fe581320699. The later deltas were checked: fe581320699 changes import paths/merges only; 3ec6b1eeb72 changes test/type infrastructure and an unused parameter; 9302c2bd510 changes popover menu layering/locales and equivalent folder-catalog argument plumbing; 06d27575c21 changes only the launch placeholder CSS and its assertions. The final panel read the current mechanisms and accepted their labelled differences, including the narrow group-hook removal. These changes introduce no different draft authority or lifecycle mechanism.

Latest main merge

The main-fix merge 4dd740fc19b6facbb1b6e62d73d3a72d011fb898 brings in #37370 and the other intervening main updates. The existing merge-check reviewer checked both conflict resolutions against both parents and affected callers and found no merge-resolution defect. Main’s agent-reference serialization is preserved beside this PR’s capture/consume/recovery, using equivalent synchronous reads of the same draft. The test retains main’s production handle and this PR’s edit-session ref; the auto-merged goal, paste and catalog paths remain intact. The owner reports six named test files passed (129/129) through tc-gate and the merge hook passed. The dictation setup screenshot remains at 5e26f380f79684146e52a85d78ece17823b721d3; a main merge preserves that evidence. No new app run was performed. The subsequent test-only commit 2bd0c872e5134f6ed2610edd10aa2f7002a43835 adds the missing agent-reference method to one test fixture, matching main’s expanded handle interface. It changes no assertions or shipped code. The same reviewer accepted the CI repair; the named test passed 3/3 and the standard hook passed. This non-shipped change also preserves QA and reference evidence.

Evidence carry

The first main sync was b94be72f9960e15dd45063fe66d1f33a0154f2d9. A later conflict blocking merge required the permitted second main merge, 8b2bea29cbc390fef3be4a8c9fcc4a44c512581a (main parent 54b496cef10454787663edf47e8a13d4f68577a4). The continued readiness reviewer narrowly checked both parents and affected callers and found no merge-resolution defect: main’s deferred-removal handoff does not reach setup scopes, append’s position/independent arguments remain distinct, and the subscribed picker retains workspace/host setup keys. The owner’s tc-gate merge run reports 26 files / 309 tests passing; the saved log does not enumerate the file names. The merge review was source-only; no additional app run was performed. Main’s new /goal chip also reaches this setup composer; its setup interaction was source-checked, without a dedicated rendered test. The inherited main lint failure then required another permitted main merge, f1509e5f89cd0265b291a222286a3970e0a30953 (main parent fbe891ce8b50bd810f9e1b1bfb2b273af846a676), bringing in #37314. It merged without conflicts. The merge delta equals main’s seven commits exactly; the only overlap with this PR is four unrelated English translation keys, whose semantic delta matches main. The earlier CLEAN narrow merge check carries with no new reviewer or round. The later commit a013c9d31622fa626f8d47eddccf60af14fc9fc9 changes only three promise type annotations in use-native-chat-composer-submit.test.tsx to match main’s boolean send result. The existing continued reviewer verified the four CI errors against the actual repair: no assertions, runtime statements or shipped code changed, and coverage is unchanged. The owner reports its named tc-gate test passed 27/27. This test-only commit preserves reference and screenshot evidence. The existing main plugin-test failures then required the clean main merge fc93c63a554e1a133a2c2dbb68958b743eb7be30, bringing in the main fixes from #37324 and #37326. Its delta is exactly main’s nine intervening commits; the only overlap with this PR is 11 locale JSONs. All locales parse, main’s semantic changes are preserved, and every PR-added locale change remains intact. No incoming change touches the launch/composer/draft code paths, so the prior CLEAN narrow merge review and reference/UI evidence carry. The remaining inherited Electron probe timeout required the clean main merge faa7d0d18fc8626e4ad5b99e2282e7c447d0fbb5, bringing in #37329. Its delta equals main’s two intervening commits across 88 paths with zero overlap with PR files. Every incoming tree entry matches main and every other entry matches the PR parent; none touches the launch/composer code paths. The prior CLEAN narrow review and reference/UI evidence carry. Main merges preserve the existing QA evidence; the historical UI evidence remains at 9302c2bd5106e1428ee44d190d8006d64fa7f961. macOS screenshots at fe58132069974af27e47d145476cde487bde28ff cover empty English layout, focus/Attach visibility, typed growth/scroll, per-workspace/agent retention, search option slots and light theme. 3ec6b1ee leaves those runtime flows unchanged. At 9302c2bd5106e1428ee44d190d8006d64fa7f961, targeted screenshots and ordinary mouse clicks verified the repaired / and @ lists and inner-menu Escape; the translated empty hint failed. Before the dictation addition, the only later non-merge shipped change, 06d27575c21, affects that empty-hint layout plus tests (a tests-row UI flow); retained screenshots certify the unchanged flows only. Exact-head Windows UI measurements found the hint still overflows, as disclosed below. Windows evidence is the saved lead report, not an uploaded screenshot, and the localized step is not a PASS. The current requested dictation addition is another tests-row UI change. Its single new setup-panel screenshot moves qa_head to 5e26f380f79684146e52a85d78ece17823b721d3 for that flow; earlier screenshots remain historical evidence for unchanged flows, and the localized placeholder failure stays deferred. The existing dictation hook, input source and launch submission mechanism are unchanged, so the whole-mechanism reference comparison carries across this presentation change.

Deferred follow-ups

  • P2, QA-E / RV6-1 — STA-13245: the empty translated placeholder overflows sideways. On Windows at 06d27575c2180450bc7265c8e1991955207f3b2f, French measures 429.2 px inside a 360 px paragraph (scrollWidth 437 versus clientWidth 376), shows a horizontal scrollbar and clips at “…, @ pour le”. German overruns the padding by 2.75 px. Typing removes the hint; both Launch buttons remain visible and usable. The panel deferred this one canonical defect; its repair is not claimed successful.
  • P3 — STA-13404: when focus leaves the composer with a / or @ list open, Escape can keep the setup panel open. Confirmed and deferred under the coordinator’s round-six disposition; no Escape repair is included.
  • P3 — STA-13230: Spanish says “skills” while the list heading says “Habilidades”.
  • P3, existing label behavior: a search row may still say “High” while its own retained setup shows Low. The changed option remains correct in the setup panel.

Existing main issues and coverage limits

  • STA-13130: inherited same-ID workspace/host-twin @ file lookup limitation, temporary and not repaired here.
  • STA-13228: on a fresh profile, model/effort search rows wait for the structured host's model catalog.
  • RV4-C1: existing server-side SSH reassignment tests lack a folder case. This is a main coverage concern, not a demonstrated PR regression. Renderer folder-catalog continuity is covered.

For Brennan's review

The screenshots are whole-window branch captures. The single current-head dictation screenshot is first; the older captures remain labelled historical evidence, not proof that the old focused layout still ships. No main screenshot comparison was run for this additive UI. The localization image is explicitly historical failed evidence.

Flow Evidence
Current — setup prompt with microphone (5e26f38) Launch prompt with existing dictation button
1 — Empty English compact box (fe58132) launch-empty.png
2 — Historical focus/Attach layout before dictation (fe58132) launch-focused.png
3 — Typed growth/scroll (fe58132) launch-grown.png
4 — Clickable skills list over setup (9302c2b) launch-skills.png
5 — Clickable file list over setup (9302c2b) launch-files.png
6 — Workspace/agent retention (fe58132) launch-retained.png
7 — Search preset option retention (fe58132) launch-search-slot.png
8a — Light theme (fe58132) launch-theme.png
8b — German empty-hint clipping before final CSS attempt; FAIL, not current-head proof (9302c2b) launch-localized.png

Three places for human judgment:

Review thread dispositions

  • Escape after focus leaves an open list: confirmed P3, deferred to STA-13404; disposition reply. Resolved.
  • SSH route-key encoding: refuted. Re-adoption old IDs come from removal tombstones, which exclude runtime-owned targets; both manual and imported targets receive generated ssh-${Date.now()}-${base36} IDs. Their raw and canonical keys are identical. Producer-based reply. Resolved.
  • Caller-prompt ownership: refuted by the current callers. Notes use the menu without picker entries and call runLaunch(row) without a setup draft; the setup/tab-bar hook supplies no caller prompt. The alleged caller-prompt/setup-draft combination is not reachable. Resolved.

Current PR diff (main merge base 81ec82b3651703f6ab3ba1a74b729453919e69e5): production +1,715/−441 in 73 files; tests +4,457/−26 in 37 files.

Review is complete on 2bd0c872e5134f6ed2610edd10aa2f7002a43835. The required ruling check passed (eight design/repeat findings, seven rulings). QA carries from 5e26f380f79684146e52a85d78ece17823b721d3 across the main merge and non-shipped test repair, as detailed above. The accepted follow-ups remain disclosed. No PR merge was performed.

@brennanb2025

Copy link
Copy Markdown
Contributor Author

The agent setup prompt now shows the dictation mic, to the right of Attach, with the same toggle and hold behavior as the chat composer. Here it's greyed out because dictation isn't set up in this test profile.

launch-dictation-empty.png

@brennanb2025
brennanb2025 merged commit 10a623b into main Nov 2, 2026
58 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant