loop

Run a prompt or slash command on a recurring interval. The interval timer starts after the previous iteration completes (post-completion delay), so runs never overlap even when execution takes longer than the interval.

Usage

/loop 5m check the deploy status
/loop 30s /stats
/loop 1h summarize git log --oneline -20
/loop stop
/loop status

Subcommand	Behavior
<interval> <prompt>	Start (or replace) a recurring loop
stop / off	Cancel the active loop
status	Show current loop state
(empty)	Show status or usage

Interval format

Accepts <number><unit> where unit is s (seconds), m (minutes), or h (hours). Examples: 30s, 5m, 1h.

How it works

1. /loop 5m check deploy parses the interval and prompt, then schedules a timer.
2. When the timer fires, the prompt is sent via sendUserMessage — triggering a normal agent turn.
3. After the agent completes (agent_end event), the next timer is scheduled with the configured delay.
4. The status bar shows a live countdown during wait periods and a “running” indicator during execution.

Behavior

• One loop at a time — Starting a new loop replaces any active loop.
• Post-completion delay — The interval is the quiet time between iterations, not the wall-clock period.
• Slash commands as prompts — /loop 5m /stats works — the prompt goes through the full input pipeline including command expansion.
• Session lifecycle — Loops are automatically stopped on session switch, shutdown, or /clear.

Extension metadata

Field	Value
Category	automation
Requires	(none)