update

3 files changed, 18 insertions, 6 deletions

diff --git a/pi/agent/extensions/loop-scheduler/README.md b/pi/agent/extensions/loop-scheduler/README.md
index 2f7a7cc..3e7ef37 100644
--- a/pi/agent/extensions/loop-scheduler/README.md
+++ b/pi/agent/extensions/loop-scheduler/README.md
@@ -5,6 +5,7 @@ Session-scoped recurring and reactive prompts for Pi.
 This extension adds two commands for interactive Pi sessions:
 
 - `/loop` re-sends a prompt on an interval while the current Pi process stays open.
+  The first run fires right after you create the loop; later runs use the interval.
 - `/watch` posts a predefined prompt when the agent becomes idle or when an assistant response contains a substring.
 
 ## Commands
@@ -66,8 +67,8 @@ Start Pi in the repo, then run:
 /loop 10m check whether the deployment finished and summarize what changed
 ```
 
-Pi will keep re-injecting that prompt every 10 minutes while the session stays
-open.
+The first run happens almost immediately (then every 10 minutes) while the
+session stays open.
 
 ### Flow 2: Loop another command
 
@@ -140,6 +141,7 @@ Preset lines use:
 
 Loop jobs do not spam turns while Pi is busy.
 
+- for **Gemma 4** models, loop/watch prompts use `deliverAs: "followUp"` so Pi queues them if the runtime is still finishing a turn (avoids “Agent is already processing” after long streams)
 - if a job becomes due while the agent is running, it is marked pending
 - when the current work finishes, the next pending loop fires once
 - missed intervals do not stack into a catch-up storm
diff --git a/pi/agent/extensions/loop-scheduler/index.ts b/pi/agent/extensions/loop-scheduler/index.ts
index b6ed621..af6b003 100644
--- a/pi/agent/extensions/loop-scheduler/index.ts
+++ b/pi/agent/extensions/loop-scheduler/index.ts
@@ -377,6 +377,13 @@ export default function loopSchedulerExtension(pi: ExtensionAPI): void {
 		lastCtx = ctx;
 	}
 
+	/** Gemma 4 sessions often hit a short "still processing" window after a turn; Pi requires deliverAs then. */
+	function sendScheduledUserMessage(prompt: string): void {
+		const id = lastCtx?.model?.id ?? "";
+		const gemma4 = /gemma[-_]?4|gemma4/i.test(id);
+		pi.sendUserMessage(prompt, gemma4 ? { deliverAs: "followUp" } : undefined);
+	}
+
 	async function openPresetFile(ctx: ExtensionContext, filePath: string, template: string, errorPrefix: string): Promise<void> {
 		if (!existsSync(filePath)) {
 			try {
@@ -537,7 +544,7 @@ export default function loopSchedulerExtension(pi: ExtensionAPI): void {
 		updateUi();
 
 		try {
-			pi.sendUserMessage(job.prompt);
+			sendScheduledUserMessage(job.prompt);
 			notify(`Loop ${job.id} fired (${reason}).`, "info");
 		} catch (error) {
 			agentBusy = false;
@@ -600,7 +607,7 @@ export default function loopSchedulerExtension(pi: ExtensionAPI): void {
 		updateUi();
 
 		try {
-			pi.sendUserMessage(job.prompt);
+			sendScheduledUserMessage(job.prompt);
 			notify(`Watch ${job.id} fired (${reason}).`, "info");
 		} catch (error) {
 			agentBusy = false;
@@ -656,7 +663,8 @@ export default function loopSchedulerExtension(pi: ExtensionAPI): void {
 			intervalMs,
 			intervalLabel,
 			createdAt: Date.now(),
-			nextRunAt: Date.now() + intervalMs,
+			// First fire is ASAP; handleJobDue then sets nextRunAt to now + intervalMs for the repeating cadence.
+			nextRunAt: Date.now(),
 			pending: false,
 			paused: allPaused, // inherit the current global pause state so new jobs added while paused start paused
 			runs: 0,
diff --git a/pi/agent/extensions/loop-scheduler/loop-presets.md b/pi/agent/extensions/loop-scheduler/loop-presets.md
index 8d141fe..f90cad3 100644
--- a/pi/agent/extensions/loop-scheduler/loop-presets.md
+++ b/pi/agent/extensions/loop-scheduler/loop-presets.md
@@ -8,4 +8,6 @@
 # * monitor: 10m check if there are any errors in the logs
 
 * tasks: 1m automatically start with the next task with fresh context if the current task completed following the agent-task-management skill.
-* proceed: 1m proceed
+* proceed: 1m proceed with the next task following agent-task-management if the previous or currently tasks being worked on is completed and committed to git.
+* review: 1m review all code changes since the last review and add code review comments using agent-task-management skill. use go-bestpractices and SOLID skills.
+* scifi: 1m write a scifi story about the current project or continue writing the story into STORY.md.