test: harden qa eval scenarios

extensions/qa-lab/src/model-switch-eval.test.ts

@@ -18,6 +18,22 @@ describe("qa model-switch evaluation", () => {
     ).toBe(true);
   });
 
+  it("accepts concise kickoff note confirmations", () => {
+    expect(
+      hasModelSwitchContinuityEvidence(
+        "Handoff clean: after the model switch, I reread the kickoff note.",
+      ),
+    ).toBe(true);
+  });
+
+  it("accepts concise paraphrases of the kickoff task after a handoff", () => {
+    expect(
+      hasModelSwitchContinuityEvidence(
+        "Handoff is clear: after the model switch, read source and docs first, run seeded qa-channel scenarios, and report worked, failed, blocked, and follow-up.",
+      ),
+    ).toBe(true);
+  });
+
   it("rejects unrelated handoff chatter that never confirms the kickoff reread", () => {
     expect(
       hasModelSwitchContinuityEvidence(

extensions/qa-lab/src/model-switch-eval.ts

@@ -7,7 +7,13 @@ export function hasModelSwitchContinuityEvidence(text: string) {
   const mentionsKickoffTask =
     lower.includes("qa_kickoff_task") ||
     lower.includes("kickoff task") ||
-    lower.includes("qa mission");
+    lower.includes("kickoff note") ||
+    lower.includes("qa mission") ||
+    (lower.includes("source and docs") &&
+      lower.includes("qa-channel scenarios") &&
+      lower.includes("worked") &&
+      lower.includes("blocked") &&
+      lower.includes("follow-up"));
   const hasScopeLeak =
     lower.includes("subagent-handoff") ||
     lower.includes("delegated task") ||

extensions/qa-lab/src/scenario-catalog.test.ts

@@ -38,6 +38,9 @@ describe("qa scenario catalog", () => {
     const discovery = readQaScenarioById("source-docs-discovery-report");
     const discoveryConfig = readQaScenarioExecutionConfig("source-docs-discovery-report");
     const fallbackConfig = readQaScenarioExecutionConfig("memory-failure-fallback");
+    const fanoutConfig = readQaScenarioExecutionConfig("subagent-fanout-synthesis") as
+      | { expectedReplyGroups?: unknown[][] }
+      | undefined;
 
     expect(discovery.title).toBe("Source and docs discovery report");
     expect((discoveryConfig?.requiredFiles as string[] | undefined)?.[0]).toBe(
@@ -46,6 +49,8 @@ describe("qa scenario catalog", () => {
     expect(fallbackConfig?.gracefulFallbackAny as string[] | undefined).toContain(
       "will not reveal",
     );
+    expect(fanoutConfig?.expectedReplyGroups?.flat()).toContain("subagent-1: ok");
+    expect(fanoutConfig?.expectedReplyGroups?.flat()).toContain("subagent-2: ok");
   });
 
   it("keeps the character eval scenario natural and task-shaped", () => {