/kill slash command silently does nothing

[Rustam-Z]

Summary

The owner-only /kill slash command is non-functional. Sending it produces no visible effect — no shutdown, no error message, no acknowledgement.

Expected

Per the documented harness behavior, /kill should trigger a graceful shutdown of the bot process. At minimum it should:

• Acknowledge receipt in the chat (or visibly fail with a reason).
• Stop the dispatcher / CC subprocess.

Actual

Nothing happens. The bot keeps running. No error surfaces in the chat. Unclear whether the command is being intercepted by the harness at all, or whether the handler silently early-returns.

Steps to reproduce

1. From the owner's account, in a DM with a bot running on pyclaudir, send /kill.
2. Observe no shutdown, no reply, no log message in the chat surface.

Suggested investigation

• Verify the slash-command interceptor is wired and not shadowed by another handler.
• Confirm the owner-check path: is the command being matched but failing the owner gate silently? If so, surface a "not authorized" reply for non-owners and an explicit "shutting down" reply for the owner.
• Check whether the kill signal is sent but the process ignores SIGTERM and there's no SIGKILL fallback.
• Add an end-to-end test: owner sends /kill → expect process exit within N seconds.

Severity

Medium — it's an availability/ops issue. Owner currently has no clean way to stop a misbehaving bot from inside Telegram and must SSH into the host. Pairs with #39 (auto-restart on wedge): if /kill worked, manual recovery from a wedged session would be one message instead of a server login.

Notes

Filed at owner's request after observing the wedge incident with @nodir_claude_bot earlier today.