Collapsible worker/background agent panel showing active sub-agents, model, reasoning level, and task description

[berrat]

Describe the feature or problem you'd like to solve

When the main agent spawns background or worker sub-agents (e.g. via the task tool — explore, task, general-purpose, code-review), there is no at-a-glance visibility into what is running. You can't see how many sub-agents are active, which model they're using, what reasoning level they're at, or what they're working on — without waiting for them to finish and report back. For complex sessions with multiple parallel agents, this is a blind spot.

Proposed solution

Add an optional Worker Agents panel in the TUI — collapsible/expandable — that shows a live list of all active background and worker sub-agents. The panel has two states:

Folded (compact, default on):

⚙ Workers  [3 active]

Unfolded (expanded):

⚙ Workers  [3 active]
  ● explore         claude-haiku-4.5   —  "Finding auth-related files in src/"
  ● task            claude-haiku-4.5   —  "Running test suite"
  ● general-purpose claude-sonnet-4.5  high  —  "Refactoring login handler"

Each row shows:

• Status indicator — spinning/active (●), done (✔), failed (✖)
• Agent type — explore, task, general-purpose, code-review
• Model — e.g. claude-sonnet-4.5, claude-haiku-4.5
• Reasoning level (optional, shown only when non-default) — e.g. high
• Task description (optional, truncated) — the short description passed to the agent

Controls:

• Toggle fold/unfold: dedicated keyboard shortcut (e.g. Ctrl+W or similar — subject to terminal keybinding availability) or by clicking the panel header
• Panel can be fully disabled via config ("workerPanel": false in config.json) for users who prefer a minimal UI
• Completed agents fade/clear from the panel after a short delay (configurable)

Example prompts or workflows

1. Main agent launches 3 parallel explore sub-agents. User sees ⚙ Workers [3 active] badge immediately, expands to see what each is scanning.
2. A task agent running tests fails. Panel shows ✖ task claude-haiku-4.5 — "Running test suite" in red — user knows before the main agent reports back.
3. User prefers a clean UI — sets "workerPanel": false in config.json, panel never appears.
4. Long-running general-purpose agent with high reasoning — user can confirm the right reasoning level is being used without having to wait for the result.
5. User is on a metered plan and wants to monitor which models are being invoked in real time.

Additional context

• OS: Windows 11 Pro (Build 26200 / 24H2)
• Shell: PowerShell 7.5.4
• Copilot CLI: 0.0.420 (win32-x64)
• The existing TUI already renders collapsible alt-screen panels (checkpoints, context, MCP) — this panel follows the same pattern.
• config.json already carries reasoning_effort — surfacing per-agent reasoning level in the panel is a natural extension.
• Model names shown in examples reflect the bundled agent defaults (explore and task use claude-haiku-4.5; general-purpose and code-review use claude-sonnet-4.5). The panel should reflect whatever model is actually configured or overridden at runtime.