[FEATURE] Agent transparency: status dashboard and live output for background agents

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

When using the Agent tool with run_in_background: true, there is no visibility into what background agents are doing. As a user, I have no way to:

1. See agent status (running / waiting for tool result / writing files / idle)
2. See which files an agent is reading or modifying in real-time
3. See the agent's reasoning or progress before it completes
4. Get push notifications when an agent finishes — the main conversation only learns about completion when the user sends their next message

This is especially concerning when agents have write access (bypassPermissions) and are modifying files in the repository. The user has to trust blindly that the agent is doing the right thing.

When orchestrating teams of 3-4 agents working in parallel, each running for 10-30 minutes, there is zero visibility into their progress.

Proposed Solution

An Agent Status Dashboard accessible from the main conversation, showing:

1. Status per agent: name, status (running/completed/failed/killed), elapsed time
2. Live activity stream: which tools the agent is calling (Read, Write, Edit, Bash), on which files
3. File change indicators: highlight when an agent modifies a file, so the user can review
4. Completion notifications: proactive notification in the terminal when an agent finishes, without requiring the user to send a message first
5. Progress summary: a brief, auto-updating status line (e.g. "Agent team-alpha: processing chapter 6/7, last action: Write 06-selbstbewusst.optimized.md")

This could be implemented as:

• A /agents command that shows a live-updating table (similar to docker ps)
• A status bar line in the terminal showing active agent count and last action
• Optional: a side panel in IDE integrations (VS Code, JetBrains)

Alternative Solutions

Currently I work around this by:

1. Manually reading the agent's output file via tail on the raw output path — but this shows raw JSON with tool calls, not human-readable progress
2. Asking Claude to check by sending a message like "?" — but this only works reactively and wastes a conversation turn
3. Using git status/git diff after agents complete to see what changed — but this is post-hoc, not real-time

None of these provide the real-time transparency needed when agents have write access to your files.

Priority

High - Significant impact on productivity

Feature Category

Interactive mode (TUI)

Use Case Example

Example scenario:

1. I launch 4 background agents in parallel, each processing different files
2. Each agent reads source files, applies transformation rules, writes output files, and validates with shell scripts
3. Agents run for 10-30 minutes each
4. During this time, I want to continue working with the main conversation
5. Problem: I have no idea what the agents are doing. Are they stuck? Have they written broken files? Did one fail silently?
6. With this feature: I could glance at a status dashboard, see which files are being modified, and know everything is on track
7. When an agent completes, I'd get a notification immediately instead of discovering it on my next message

Additional Context

The /agents command exists but only shows a static list — it doesn't provide live progress or activity streams.

Related concern: when agents run with bypassPermissions, users are granting significant trust. Transparency into what the agent is actually doing with that trust would make the feature much more comfortable to use.