Spike: evaluate `amq monitor` as a Claude Code watcher instead of `amq wake`

[avivsinai]

Problem

Today the main real-time Claude Code pattern is still amq wake --me claude &, which depends on a background Go process plus TIOCSTI terminal injection. That works, but it is explicitly experimental and can fail on hardened systems.

We now have amq monitor, a combined watch+drain command that emits structured output and is documented as suitable for background watcher workflows. We should check whether it gives us a better Claude Code integration surface than running wake.

Question

Can amq monitor replace or reduce our need for amq wake in Claude Code workflows?

Specifically:

• does amq monitor --include-body provide a practical inbox-awareness loop for Claude Code without relying on TIOCSTI?
• what is the cleanest UX: separate terminal pane, tmux pane, notify bridge, statusline helper, or something else?
• does it improve portability / reliability enough to recommend over wake for Claude Code users?
• what are the tradeoffs on latency, noise, and message handling semantics (auto-drain vs peek)?

Deliverable

A small spike that compares wake vs monitor for Claude Code and ends with one of:

• recommend monitor as the new default pattern
• recommend monitor only in specific environments (for example when TIOCSTI is unavailable)
• reject monitor as a Claude Code replacement and document why

Notes

• This is about Claude Code UX only.
• No need to redesign wake or invent a new daemon first.
• Reuse the current amq monitor command as-is for the spike.