fix(chat): keep agent activity visible between actions

[maria-rcks]

Running turns now keep an animated Thinking row at the tail whenever no tool is active, including after commentary and completed tools. Terminal tool details remain visible but stop shimmering once completed. Verified with 87 focused web tests, web typecheck, lint, formatting, and a real showcase browser pass; screenshot and video capture were unavailable in the shared preview client.

Built with gpt-5.6-sol in the Codex harness.

---

Note

Low Risk
Changes are confined to chat timeline row derivation and rendering in the web app, with no auth or data-path impact; the main caveat is the MessagesTimelineRow shape change if anything outside this PR pattern-matches on working.showThinking.

Overview
Running turns now show Thinking as its own timeline row whenever no tool is actively running, instead of tucking it under the Working for… header via showThinking. That row stays at the tail through gaps after commentary, completed tools, or a fresh send before a turn id exists.

work-live rows gain an active flag: in-progress tools still use the animated live activity line and Running labels; once the latest tool finishes (or is declined) but the turn is still running, the same row stays visible with static styling and past-tense copy (Ran / Declined). A separate thinking row appears when there is no active live tool row.

Timeline derivation picks the displayed tool from the latest still-active entry when present, and stable row reuse keeps the thinking row instance stable as assistant text streams in.

^{Reviewed by Cursor Bugbot for commit 60fb030. Bugbot is set up for automated code reviews on this repo. Configure here.}

Note

Split agent 'thinking' indicator into its own timeline row in MessagesTimeline

• Adds a dedicated thinking row variant to types.MessagesTimelineRow, replacing the showThinking flag previously embedded in the working variant
• Adds an active: boolean field to the work-live variant so deriveMessagesTimelineRows can distinguish actively running tool groups from completed ones; active groups render as live rows while inactive ones render as static content blocks
• Updates liveWorkEntryLabel to return Running, Declined, or Ran based on the new active flag and lifecycle status, e.g. Running pnpm vs Declined pnpm\n- Updates isRowUnchanged to compare thinking rows by createdAt (like working rows) and to detect changes in the active flag for work-live rows
• Behavioral Change: WorkingTimelineRow no longer renders thinking content; callers relying on the working row's showThinking field or embedded thinking markup will break. The new ThinkingTimelineRow component handles all thinking rendering.

^{Macroscope summarized 60fb030.}

[cursor]

Cursor Bugbot has reviewed your changes using high effort and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit f1bc202. Configure here.}

[macroscopeapp]

Approvability

Verdict: Approved at 60fb030

Macroscope's review found this PR approvable — This is a focused chat-timeline rendering fix that keeps completed activity visible and adds a stable Thinking indicator between actions. Changes are isolated to the web timeline derivation/rendering logic with targeted tests and no broader data, API, security, or infrastructure impact.

^{You can add or adjust custom eligibility rules. Learn more.}