[Bug]: 2.2.7 removed the only live indicator that a command or agent is still running (StatusPanel deleted with the todo list)

[KayCe75]

Summary

2.2.7 removed the only live indication that a command or background agent is still running. There is now no way to tell "still working" from "finished" while a turn is in progress.

What was removed

PR #1301 ("retire task-list UI and disable Claude task tools", merged 2026-09-10) deleted src/features/chat/ui/StatusPanel.ts in full (286 lines). Its own header comment described it as:

StatusPanel - persistent bottom panel for todos and command output.

And its data model was explicitly about running commands, not about todos:

export interface PanelBashOutput {
  id: string;
  command: string;
  status: 'running' | 'completed' | 'error';
  output: string;
  exitCode?: number;
}

const MAX_BASH_OUTPUTS = 50;

The PR rationale addresses only the todo list:

Current models are capable enough to complete ordinary tasks without maintaining a separate tool-owned todo list. The task panel and Claude task-list tools add UI, state, and execution bookkeeping without materially improving task performance.

That reasoning may well hold for the todo list. But the same file also carried the live command/output section, and that part was removed along with it without being discussed. Two different features shared one panel; only one of them was actually evaluated.

Why this matters in practice

Without it, the user cannot tell whether the assistant is still working:

• A long-running script is started. The panel used to show it as running.
• Now nothing is shown. The chat looks idle.
• The user assumes the turn is finished, asks "where are the results?", and interrupts work that was still in progress.

This is not hypothetical — it is exactly what happened here today, repeatedly, after Obsidian was restarted and 2.2.7 became active. Tasks that take minutes (background agents, long shell commands, multi-file analysis) are now indistinguishable from a stalled session.

Claude Code's own spinner wording ("Cooking…", "Philosophising…") is still visible, but it only signals that something is happening, never what. With several parallel agents it says nothing useful at all.

Interaction with #1338

This compounds with #1338 in the same release. Together:

• During the turn: no indication of what is running (this issue)
• After the turn: tool calls and any assistant text preceding them are collapsed into Worked for MM:SS ([Bug]: 2.2.7 collapses assistant text written before a tool call into the "Worked for …" block #1338)

So the user has no visibility into the work while it happens, and none afterwards either without hunting for a collapsed header. Before 2.2.7 both were visible by default.

Suggested resolution

Not a request to bring back the todo list. Only the command-activity part:

1. Restore the PanelBashOutput section of the panel (running / completed / error with the command line), without the todo section; or
2. Provide a minimal persistent indicator while a turn is active — e.g. a single line showing the currently running tool and, if applicable, the number of active background agents; or
3. Keep tool call elements expanded and visible while they are running, and collapse them only once the turn completes.

Any of the three would restore the distinction between "still working" and "done", which is currently unavailable.

Environment

• Claudian 2.2.7 (installed 2026-09-12, active since Obsidian restart 2026-09-14 17:50)
• Obsidian 1.13.7, Linux (NixOS, AppImage via appimage-run)
• Claude Code CLI 2.1.263
• Default theme, no CSS snippets

[KayCe75]

Clarification of the exact failure case, because my original description was too broad.

Two indicators do still work and are not the problem:

1. Claude Code's own spinner with the elapsed-time counter ("Bear with me… (esc to interrupt) 14s"). Visible for as long as the turn is active.
2. Tool call rows in the transcript. A Bash call renders as a single expandable line showing the command; clicking it reveals (a truncated view of) the output.

The gap is specifically work that continues after the turn has ended:

• The assistant starts a background command or a background agent.
• That tool call is marked completed — because dispatching it succeeded.
• The assistant writes "I'll report back when it's done" and the turn ends.
• The spinner stops. The elapsed-time counter disappears.
• Nothing anywhere indicates that a process or agent is still running.

To the user the session now looks finished, and the natural reaction is to ask "so where are the results?" while the work is still in flight — or worse, to start something else on top of it.

This is the case the deleted StatusPanel covered, and it is why it was described as persistent: its PanelBashOutput entries carried status: 'running' | 'completed' | 'error' and survived across turns, so a long-running command stayed visible as running after the assistant had stopped emitting output.

Today there is no equivalent. The spinner is bound to turn activity, and the tool call row shows dispatch success, not process state.

So the minimal fix is narrower than what I suggested initially — the todo section is genuinely not needed. What is needed is a single persistent line while any background command or agent is still running, e.g.:

2 background tasks running — scan1.py, agent review:bugs

Visible regardless of whether a turn is currently active, disappearing when the last one finishes.

[github-actions]

This issue has been inactive for 7 days and will be closed in 7 days if no further activity occurs.