fix(auto-reply): require durable restart recovery

vincentkoc · vincentkoc · commit 96a96b7e2bad · 2026-06-22T13:56:25.000+08:00
diff --git a/src/auto-reply/reply/agent-runner-execution.test.ts b/src/auto-reply/reply/agent-runner-execution.test.ts
@@ -6031,6 +6031,7 @@ describe("runAgentTurnWithFallback", () => {
       sessionKey: "main",
       getActiveSessionEntry: () => undefined,
       resolvedVerboseLevel: "off",
+      isRestartRecoveryArmed: () => true,
     });
 
     expect(result.kind).toBe("final");
@@ -6085,6 +6086,7 @@ describe("runAgentTurnWithFallback", () => {
       sessionKey: "main",
       getActiveSessionEntry: () => undefined,
       resolvedVerboseLevel: "off",
+      isRestartRecoveryArmed: () => true,
     });
 
     expect(result).toEqual({
diff --git a/src/auto-reply/reply/agent-runner-execution.ts b/src/auto-reply/reply/agent-runner-execution.ts
@@ -1594,6 +1594,7 @@ export async function runAgentTurnWithFallback(params: {
   toolProgressDetail?: "explain" | "raw";
   replyMediaContext?: ReplyMediaContext;
   onCompactionNoticePayload?: (payload: ReplyPayload) => Promise<void> | void;
+  isRestartRecoveryArmed?: () => boolean;
 }): Promise<AgentRunLoopResult> {
   const TRANSIENT_HTTP_RETRY_DELAY_MS = 2_500;
   let didLogHeartbeatStrip = false;
@@ -3117,6 +3118,14 @@ export async function runAgentTurnWithFallback(params: {
       // bookkeeping for drain/lane-cleared is still recorded.
       if (isReplyOperationRestartAbort(params.replyOperation)) {
         takePendingLifecycleTerminal()?.emit("end", err);
+        if (params.isRestartRecoveryArmed?.() !== true) {
+          return {
+            kind: "final",
+            payload: markAgentRunFailureReplyPayload({
+              text: buildRestartLifecycleReplyText(),
+            }),
+          };
+        }
         return {
           kind: "final",
           payload: {
diff --git a/src/auto-reply/reply/agent-runner.ts b/src/auto-reply/reply/agent-runner.ts
@@ -1568,6 +1568,18 @@ export async function runReplyAgent(params: {
       }
     }
   };
+  const isRestartRecoveryArmed = (): boolean => {
+    if (!trackedRestartRecoveryDeliveryContext || !sessionKey || !storePath) {
+      return false;
+    }
+    const persisted = loadSessionEntry({
+      sessionKey,
+      storePath,
+      clone: false,
+      hydrateSkillPromptRefs: false,
+    });
+    return persisted?.abortedLastRun === true || activeSessionEntry?.abortedLastRun === true;
+  };
   const prePreflightCompactionCount = activeSessionEntry?.compactionCount ?? 0;
   let preflightCompactionApplied;
 
@@ -1742,6 +1754,7 @@ export async function runReplyAgent(params: {
         resolvedVerboseLevel,
         toolProgressDetail,
         replyMediaContext,
+        isRestartRecoveryArmed,
       }),
     );
 
@@ -2596,9 +2609,25 @@ export async function runReplyAgent(params: {
     // terminal: it looks like the owed work was abandoned and invites a
     // duplicate manual retry. `aborted_for_restart` is an "aborted" result, so
     // it falls through to the shared abort branch below.
-    if (replyOperation.result?.kind === "aborted") {
+    if (
+      replyOperation.result?.kind === "aborted" &&
+      replyOperation.result.code === "aborted_by_user"
+    ) {
       return returnWithQueuedFollowupDrain({ text: SILENT_REPLY_TOKEN });
     }
+    if (
+      replyOperation.result?.kind === "aborted" &&
+      replyOperation.result.code === "aborted_for_restart"
+    ) {
+      if (isRestartRecoveryArmed()) {
+        return returnWithQueuedFollowupDrain({ text: SILENT_REPLY_TOKEN });
+      }
+      return returnWithQueuedFollowupDrain(
+        markReplyPayloadForSourceSuppressionDelivery({
+          text: RESTART_LIFECYCLE_REPLY_TEXT,
+        }),
+      );
+    }
     if (error instanceof GatewayDrainingError) {
       replyOperation.fail("gateway_draining", error);
       return returnWithQueuedFollowupDrain(
