[teamleaderleo]

Builds on #22675 (transcript-density's action rows, merged).

Problem

Leo (2026-10-20): the rows under chat messages should match Claude Code desktop.

• Hovering an answer drew a box around it with "6:29 PM" and Copy at the far right.
• Under a prompt there was only a lone bookmark icon, plus "Steered · the agent read it" under steers.

Change

• Under an answer (TurnRows.tsx TurnFooter), at the answer's left edge:
  • Copy, Fork from here, Pin, Read aloud, then the answer's age ("43 minutes ago"; the exact date and time on hover).
  • Fork's tooltip has Claude's second line: "Starts a new session from this point. This one stays as it is." (data-tooltip-detail in ui/titleTooltips.ts).
  • Retry moved to the prompt. The answer footer no longer draws a focus outline; focusing it shows its row instead.
• Under a prompt (UserBubble.tsx), at the bubble's right edge:
  • the age, a read receipt, Copy, Edit, Retry (only on the last ended turn's prompt), Fork and Pin.
  • The read receipt is ✓ "Delivered when the current step ends" while a steer waits (steer-default's wording; it drops its own waiting marker) and ✓✓ once the agent read it; hovering ✓✓ shows "Read by Claude Code" (the harness name).
• Hover only. Neither row shows until you hover or focus it. Each keeps its height, so nothing shifts when it appears. Hovering a prompt shows only the prompt's row, not its turn's footer too, so one row shows at a time.
• Removed: SteerMark ("Steering…" / "Steered · the agent read it"), the bookmark icon under prompts, and the always-on row for the latest prompt and turn.
• Pin reuses the thread minimap's bookmarks. acpmux now also accepts a turn's turn_result seq, so answers can be pinned too (hub/bookmarks.rs).
  • Pinned prompts show in the minimap as before.
  • Pinned answers aren't listed in the summary panel yet (follow-up).
• Read aloud uses the web view's speechSynthesis (the macOS voice). It is one message at a time, and pressing it again stops.
• New strings, in all 21 locales: message.readBy, message.readByAgent, message.delivered, message.pin, message.unpin, message.readAloud, message.stopReading, turn.forkDetail. Removed the unused steer.*, message.read and message.unread.

Observed / Not run

• Ran scripts/cmux-next/prepush-gate.py: no new violations.
• Not run locally: web lint, typecheck, vitest and cargo. Cosmic-gray was at load 80, so CI runs these. Tests that pinned the old rows still need updating (in progress).

Changelog

• Changed: Chat messages show Claude-style hover action rows: Copy, Fork, Pin, Read aloud and the message's age under answers; age, read receipt, Copy, Edit, Retry and Fork under prompts.

Agent session: cosmic-gray

🤖 Generated with Claude Code

---

Summary by cubic

Matches Claude Code desktop's hover action rows under every chat message, on hover or focus only. Answers show Copy, Fork (with Claude's "starts a new session" tooltip), Pin, Read aloud and the message's age under their left edge; prompts show age, a read receipt, Copy, Edit, Retry, Fork and Pin under the bubble's right edge; another agent's message shows age, the receipt, Copy, Retry and Pin under its card, after its Reply. A Chief-sent prompt gets the same row as one the person sends. The receipt is one check while a steer waits and two once read ("Read by "). The Steered line, the lone bookmark icon and the always-on latest rows are gone, and no answer footer draws a focus box. This also reconciles with base's steer-as-one-turn, row-height-estimate, Reply placement and rewritten Retry model menu (RetryModelMenu.tsx) changes, which the prompt's row and a failed turn's footer now reuse; Copy reuses base's CopyButton (tooltip name, resetting check).

Bug Fixes

• Fixes the row-height estimate to reserve a line under a steer only while it waits.
• A pinned prompt or answer keeps its pin through a fork.
• The waiting receipt's tooltip gets its own locale key: base's message.delivered belongs to MessageCard.

^{Written for commit 1e73560. Summary will update on new commits.}

[teamleaderleo]

Bead cx-3c3ng created for this PR by the cmux coordinator.

[github-actions]

cmux status _{(one comment, edited in place; each line is updated by the job that owns it)}

Review bdf0136: HQ feed card · gallery before/after runs · checks

App of this head (app.zip, kept 1 day)

gh workflow run reload-build.yml -R manaflow-ai/cmux --ref feat-cmux-next -f tag=pr22773 -f ref=leo/message-hover-rows -f platform=macos -f backend_mode=local

Unclear: web / react-apps-check failed: no known signature; failed step: Check webview test shards.

CI failed on bdf01361a4 (run 39566576086 attempt 1): 4 code, 1 unknown.

Job	Verdict	Why
web / react-apps-check (1)	code	a web test failed
web / react-apps-check (2)	code	a web test failed
web / react-apps-check (4)	code	a web test failed
web / react-apps-check (3)	code	a web test failed
web / react-apps-check	unknown	no known signature; failed step: Check webview test shards

Failures:

• seen on other PRs composer-claude-split.test.ts chromium: + is its glyph, named Add with no tooltip: failed (job)
• seen on other PRs composer-claude-split.test.ts webkit: + is its glyph, named Add with no tooltip: failed (job)
• seen on other PRs composer-history.test.ts ↑ and ↓ in an empty prompt step through the chat's earlier prompts: failed (job)
• seen on other PRs composer-suggested-prompt.test.tsx nothing is suggested from another folder, another worktree, a merged PR, an old chat or a busy one: failed (job)
• seen on other PRs composer-suggested-prompt.test.tsx a folder with a trailing slash is the same checkout: failed (job)
• seen on other PRs composer-suggested-prompt.test.tsx a chat that knows its branch takes no PR from another branch: failed (job)
• seen on other PRs composer-suggested-prompt.test.tsx a chat with turns gets no suggestion: failed (job)
• seen on other PRs composer-suggested-prompt.test.tsx Tab writes the suggestion without sending; Shift-Tab does not take it; Esc puts it away: failed (job)
• seen on other PRs composer-suggested-prompt.test.tsx the chat folder's PR from the header's gh read is suggested when no session reports one: failed (job)
• seen on other PRs quiet-access-chip.test.ts chromium: Full access draws in the model chip's color: failed (job)
• ... 10 more in the job logs

Matched log lines

web / react-apps-check (1): (fail) chromium: + is its glyph, named Add with no tooltip [1973.27ms]
web / react-apps-check (2): (fail) ↑ and ↓ in an empty prompt step through the chat's earlier prompts [1920.67ms]
web / react-apps-check (4): (fail) @ then letters lists the chat's matching files and Return writes the path [2016.70ms]
web / react-apps-check (3): (fail) a chat on an SSH machine > counts its running turn from the prompt and offers no machine switch mid-chat [31190.38ms]

Not re-run automatically: web / react-apps-check (1), web / react-apps-check (2), web / react-apps-check (4), web / react-apps-check (3), web / react-apps-check are not machine failures.

Written by scripts/ci/classify_failures.py (ci-failure-attribution.yml); signatures are its SIGNATURES table. A machine verdict is the runner's fault, not this PR's; yours means the failing file is one this PR changes, also red on main that main's latest full suite fails the same way, seen on other PRs that it failed on another pull request's run lately.

red: web / Web status; base red: the same failure on the base 84494d5; does not block
red: web / react-apps-check; base red: the same failure on the base 84494d5; does not block
red: web / react-apps-check (1); base red: the same failure on the base 84494d5; does not block
red: web / react-apps-check (2); base red: the same failure on the base 84494d5; does not block
red: web / react-apps-check (3); base red: the same failure on the base 84494d5; does not block
red: web / react-apps-check (4); base red: the same failure on the base 84494d5; does not block

CI fast guards failed on bdf01361a4 (https://github.com/manaflow-ai/cmux/actions/runs/39566576086). It does not block the merge; a red guard merged into main breaks it for every open PR.
The log named no failed step; see the run.

Agents: python3 scripts/ci/guard_attribution.py fix applies the mechanical fixes locally.

[blacksmith-sh]

Found 4 test failures on Blacksmith runners:

Failures

Test	View Logs
/tmp/tmp8esvkds5/summary-fail.json (live):/summary-fail.json (live):	View Logs
/tmp/tmpgxd56a2i/summary-missing.json (live):/summary-missing.json (live):	View Logs
convention violations found (ERROR lines above)./ convention violations found (ERROR lines above).	View Logs
test_every_source_and_test_directory_is_a_target (main.GraphCoversThePackage.test_e very_source_and_test_directory_is_a_target) (directory='CmuxNextHomeEngineTests')/ test_every_source_and_test_directory_is_a_target (main.GraphCoversThePackage.test_e very_source_and_test_directory_is_a_target) (directory='CmuxNextHomeEngineTests')	View Logs

^{Need help on this PR? Tag @codesmith-bot with what you need.}

[teamleaderleo]

owner: 01a15c08-3514-7430-9c71-6b7ba4fdac72

[teamleaderleo]

Visual preview of 726388f: in the maintainers' private review feed.

[teamleaderleo]

Receipt review (merge resolution cc65008..bdf0136): PASS

Checked (read-only, git only):

• git diff 84494d5c bdf01361 and git diff 90115b72 cc65008b: the same 28 files with the same stats, and their +/- lines are byte-identical (diff of -U0 hunks is empty). The PR's delta is unchanged and nothing from base's commits 90115b7..84494d5 was dropped.
• Files both sides touched: UserBubble.tsx and IncomingMessageCard.tsx. cmux NEXT composer: a pasted image is a chip only, never data: text in the prompt #23100's UserBubble changes are all present: the useMemo and inlineImages imports, shown = withoutDataUris(text, t("image.viewer")) in the bubble, and editable = liftInlineImages(row.text ?? text).text as the editor's initial. webviews: the check is green again (three warnings it counts as a red) #23144's oxlint autoFocus comment in IncomingMessageCard is kept too. Dropping CopyButton from UserBubble is right: it is the PR's own removal (MessageCopy wraps CopyButton).
• Copy: <MessageCopy text={row.text ?? text} /> copies the prompt as sent, which matches cmux NEXT composer: a pasted image is a chip only, never data: text in the prompt #23100's "Copy copies it as sent". Retry: resend = row.resend ?? row.text ?? text can hold a legacy data: image, but cmux NEXT composer: a pasted image is a chip only, never data: text in the prompt #23100's send path in direct.ts runs liftInlineImages(input, attachments) on every send, so a retry sends those as image blocks and not as base64 text.
• Base tests: composer-paste-images.test.tsx (cmux NEXT composer: a pasted image is a chip only, never data: text in the prompt #23100) renders UserBubble and asserts only on .cv-user__bubble textContent, which is still {shown}; the PR's action row sits outside the bubble. Every hook the row uses (useRowMark, useMinuteClock, the TurnActionsContext default) also works without a provider, and image.viewer still resolves to "Image" in en. No other test touching UserBubble or inlineImages changed on base.
• l10n: Localizable.xcstrings at the head parses as JSON (1430 keys). All 895 literal t("…") keys in acpmux source exist, including the ternary keys in MessageActions.tsx.
• scripts/web-lint.sh on the head: exit 0.

Not run: bun test, vite, gallery, Swift/Rust builds and tests. Test outcomes above come from reading the code, not from running it.

Agent session: cosmic-gray

[teamleaderleo]

cmux-tui spec inventory / inventory is red on 10 of 11 fresh feat-cmux-next PR heads in the last 2:00:00, so it is most likely a base red, not this PR's. The base-reds lane has it; a base merge picks up the fix once it lands.

[teamleaderleo]

Receipt review (merge resolution bdf0136..3bd6142): PASS

Head 3bd6142 = merge of bdf0136 (reviewed) and base a18aa77. Read-only, git objects only.

Checked:

1. Delta preserved, nothing from base lost. git diff a18aa77e 3bd61427 and git diff 84494d5c bdf01361 (84494d5 = merge-base of bdf0136 and a18aa77e) touch the same 28 files with the same counts (1779+/670-), and their +/- lines are byte-identical (2449 lines each, xcstrings included). Every file outside the PR's 28 is identical to base.
2. Localizable.xcstrings: parses as JSON. Base has 1467 keys, head has 1473. Every base key is in head with base's exact value, except the 3 the PR removes (message.read, message.unread, steer.landed). The 9 added keys (message.delivered, deliveredNotRead, pin, unpin, readAloud, stopReading, readBy, readByAgent, turn.forkDetail) match bdf0136's values and are each used in source. In base a18aa77e, the removed keys are used only in SteerMark.tsx (the PR deletes it) and UserBubble.tsx (the PR rewrites those lines). Head has no uses left. All 1245 literal t("...") keys in head webviews/src/agent-session/acpmux exist in the catalog (base: 1244/1244).
3. Auto-merged files: base changed App.tsx, conversation.css, model.ts and composer.gallery.ts between 84494d5 and a18aa77e. Head has no references to SteerMark, steer.landed, .cv-user__steer (it appears only in tests that assert a count of 0), .cv-user__bookmark, data-last-turn/data-last-prompt, lastTurnId/lastPromptId or usePromptRowMark in UserBubble. Every import from conversation/MessageActions (App.tsx, TurnRows.tsx, UserBubble.tsx) names something MessageActions.tsx exports. CopyButton is still its own module, which MessageActions and base callers import. retryRowId (turns.ts) and row.resend (model.ts) both exist on head. .cv-user__status is still a live class (PromptDelivery, the queued status in App.tsx). The PR's model.ts height change removed only the steer status line.

Not run: bun test, bun run check/lint, vite/gallery, cargo, swift. This is a static read of the merge only. Known base lint reds not from this PR: the run-board timeline.css text-focus ring (#22408) and web/ eslint any.

[cursor]

Bugbot needs on-demand usage enabled

Bugbot uses usage-based billing for this team and requires on-demand usage to be enabled.

A team admin can enable on-demand usage in the Cursor dashboard.

[teamleaderleo]

Receipt review (merge resolution 3bd6142..c19ca3d): FAIL

Defect: message.delivered is now in webviews/src/agent-session/acpmux/Localizable.xcstrings twice, with two different meanings.

• Base (Agent pane: message cards say how far they got; shell rows drop the leading cd #22994, 230e98f) added message.delivered = "Delivered to {to}" for MessageCard's status line (head file line ~28809).
• The PR adds message.delivered = "Delivered when the current step ends" for ReadReceipt's waiting-steer tooltip (head file line ~400).
• Git auto-merged both blocks with no conflict, so the file has a duplicate key. JSON.parse in gen-strings.mjs keeps the last one (base's). As a result MessageActions.tsx:119 t("message.delivered") (with no to) would show "Delivered to {to}" under a waiting steer, and webviews/test/user-message-actions.test.tsx:96 and :122 (which expect "Delivered when the current step ends") would fail. Xcode/xcstrings tooling may also reject the duplicate.
• Suggested fix (owner's call): rename the PR's key (for example message.deliveredAtStepEnd), use it in ReadReceipt, and remove the PR's duplicate block. Keep base's message.delivered as it is.

Checked

1. git diff 230cc72b c19ca3d3 vs git diff a18aa77e 3bd61427: the same 28 files, and textually the same hunks except icons.tsx. There the PR no longer adds its own DoubleCheck; the one DoubleCheck (base glyph, comment now covers both uses) is followed by Pin and Speaker, unchanged. MessageCard.tsx (Check/DoubleCheck/Spinner receipt), conversation.css .cv-message__receipt, App.tsx localPaths/linkItems, the ShellBlock/commandFolder and intake changes from base are all intact in the merge (none are in the PR delta).
2. Overlap with base: base's MessageCard receipt (Sending / Sent / Delivered to X) is a separate surface (agent-to-agent tool rows) from the PR's UserBubble row, and shares only the DoubleCheck icon, which is resolved correctly. The new base tests (transcript-row-message, transcript-row-command-folder, chat-environment-line, new-tab-dropped-paths, composer-attachment-matrix) don't touch UserBubble or the PR's row. transcript-row-message expects "Delivered to leoli-24", which still holds because base's duplicate wins. The collision above is the one overlap.
3. xcstrings: Python parses it, but with a duplicate key (object_pairs_hook finds only message.delivered). Every base key is present with base's value, except message.read, message.unread and steer.landed, which the PR removes. A git grep at c19ca3d finds no source or test use of those three. Every literal t("...") key in acpmux source is in the catalog. The PR's other added keys match 3bd6142.

Not run: bun test, vite, gen-strings, cargo, swift, lints. All reading was done with git at explicit SHAs. Known base lint red (#22408 run-board timeline.css) not assessed.

[teamleaderleo]

Receipt review 1e73560: PASS

Checked (read-only through git, in a separate clone):

• git diff c19ca3d3..2ac81851 is exactly 2 lines: the catalog key message.delivered is renamed to message.deliveredAtStepEnd (comment and all 21 localizations kept), and ReadReceipt in conversation/MessageActions.tsx now calls t(waiting ? "message.deliveredAtStepEnd" : "message.deliveredNotRead").
• I parsed Localizable.xcstrings at 1e73560 with an object_pairs_hook: no key appears twice. message.delivered matches base efaab9b exactly ("Delivered to {to}", used by MessageCard). message.deliveredAtStepEnd has the PR's text ("Delivered when the current step ends") in 21 locales. No base entry was changed. The only base keys missing are message.read, message.unread and steer.landed. The PR already deleted these before this fix (they are absent at c19ca3d), and nothing under webviews/src or webviews/test refers to them.
• All 953 literal t("...") keys under webviews/src/agent-session/acpmux exist in the catalog, including the ternary pairs.
• user-message-actions.test.tsx expects "Delivered when the current step ends", "Delivered", "Read by Claude Code/Codex", "Read by the agent", "Copy", "Edit" and "Fork from here". These match the en catalog values. MessageCard.tsx and transcript-row-message.test.tsx ("Delivered to leoli-24") are the same as base, and their key keeps base's value.
• The PR delta against base, git diff efaab9b3...1e735606, is identical to the earlier git diff 230cc72b...c19ca3d3 (28 files, +1774/-671) except for the two rename lines. The merge changes nothing against base outside the PR's own files, so the clean merge lost nothing from base.

Not run: bun tests, bun run check, web lint, gallery, and any build or CI check. I only read code.

[teamleaderleo]

merge-audit: #22773 merged at 1e73560 without no SHIP review receipt on the head; no risk tier label. #23291 reverts it. To keep this change, add the missing evidence for 1e73560 here; the next audit pass then closes #23291. Otherwise review and land #23291.

[github-actions]

Shipped in:

• cmux NEXT nightly 1.0.0-nightly.3958372135301 (bc661dae0e)

[teamleaderleo]

Review (subagent, correctness first): SHIP at 1e73560

This is a post-hoc independent review of the merged change (merge commit 8aa26c6). It was read-only and nothing was run. risk:low.

Nothing blocks. Notes:

• Read-aloud state is global, keyed only by row id, and is never cancelled on a session change or pagehide (readAloud.ts:6).
• A peer daemon on an older acpmux rejects turn_result marks, so the pin undoes itself silently.
• The focused answer row has no focus outline (WCAG 2.4.7). Each footer builds an Intl.DateTimeFormat on every render. Both minor.

The merge-audit revert of this PR is handled separately. See leo-workspace briefs/2026-10-20-gui-push/revert-triage.md.