CLI: show agent/thread attribution on action status rows (read/search/update/exec)

Summary

When Codex CLI shows in-turn action/status rows (for example read/search/update/exec), the UI does not show which agent/thread produced each action.

This is a blocker for multi-agent workflows where users need to verify that each role is acting on the correct code area.

Environment

• Codex CLI: 0.101.0
• Install: @openai/codex via npm and Homebrew
• OS: macOS (Apple Silicon)

Current behavior

• Status rows show action type/state (and color), but not agent identity.
• In collaboration/sub-agent flows, attribution exists in protocol/log metadata (sender_thread_id, receiver_thread_ids, etc.), but is not surfaced inline in the UI rows users monitor.

Expected behavior

For each visible action row (read/search/update/exec/patch/mcp call):

• Show origin identity, at minimum:
• agent thread id (or stable short label)
• source (agent, user_shell, unified_exec_*) where applicable
• Keep this visible both in live progress and historical row view.

Why this matters

Without per-row attribution, users cannot reliably verify role/function boundaries in multi-agent execution, which makes debugging and trust significantly harder.

Suggested UX

• Add a compact origin badge on each row, e.g. [A:ab12], [Planner], [Mapper]
• Provide a /status or statusline toggle to show/hide attribution detail
• Optional filter: show rows for active agent only vs all agents

Protocol note

The data appears to already exist in event surfaces (sender_thread_id, receiver_thread_ids, turn_id, source), so this looks like a presentation gap rather than missing backend data.