[FEATURE] Allow resuming or reading the transcript of a completed/expired subagent

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

I orchestrate long sessions with background subagents (executor pattern). A warm subagent holds valuable context (~500k tokens of repo knowledge). After it completes and some time passes, SendMessage to it fails with "No transcript found for agent ID", and the only option is to spawn a fresh agent that has to re-learn the codebase from scratch — wasting a lot of tokens.

A related UI gap in the desktop app: when a subagent is freshly spawned, I can watch its reasoning/process live in the side panel. But when an existing subagent is resumed via SendMessage, the side panel shows nothing at all — no live activity, no transcript. So resumed agents are both cheaper and completely opaque, which discourages the token-efficient pattern.

Proposed Solution

(a) let a new message resume a completed subagent from its persisted transcript for the whole session lifetime; (b) expose the old agent's transcript (even read-only) so the orchestrator can seed a replacement agent with it; (c) show resumed subagents' live activity in the desktop side panel the same way as freshly spawned ones. The transcript files seem to exist on disk under the session's tasks directory, so this looks like a lifecycle/lookup limitation rather than missing data.

Alternative Solutions

_No response_

Priority

High - Significant impact on productivity

Feature Category

Interactive mode (TUI)

Use Case Example

_No response_

Additional Context

_No response_