test(grok): reject reasoning projected as task replies

Co-Authored-By: Paperclip <noreply@paperclip.ing>

tests/runner-e2e/run-observations.ts

@@ -81,6 +81,19 @@ function record(value: unknown): Record<string, unknown> {
     : {};
 }
 
+export function reasoningProjectionFailures(events: readonly ObservableRunEvent[]): string[] {
+  const misclassified = events.filter((event) => {
+    const envelope = record(record(event.payload).prpEvent);
+    const payload = record(envelope.payload);
+    return envelope.eventType === "item.delta" &&
+      payload.kind === "agentMessage" &&
+      record(payload.update).kind === "reasoning";
+  });
+  return misclassified.length === 0 ? [] : [
+    `provider reasoning was projected as assistant text in ${misclassified.length} durable events`,
+  ];
+}
+
 export function isNonExecutingReviewFenceRun(run: ObservableRunState) {
   return (
     run.status === "cancelled" &&

tests/runner-e2e/runner.spec.ts

@@ -27,6 +27,7 @@ import {
   isOpenRouterDeepSeekHelloTerminalVariance,
   numberedPlanStepCount,
   providerSessionContinuityFailures,
+  reasoningProjectionFailures,
 } from "./run-observations.js";
 import { resolveRunnerE2ESource } from "./source.js";
 import { isValidNativePrpEnvelope } from "./native-event-envelope.js";
@@ -1732,7 +1733,9 @@ for (const execution of executions) {
       const pendingInteractions = terminal.interactions.filter(
         (interaction) => interaction.status === "pending",
       );
-      const invariantFailures: string[] = [];
+      const invariantFailures: string[] = reasoningProjectionFailures(
+        runEventsByRun.flatMap((entry) => entry.events),
+      );
       if (pendingInteractions.length > 0)
         invariantFailures.push(
           `expected no unresolved interaction; observed ${pendingInteractions.length}`,

tests/runner-e2e/support.test.ts

@@ -44,10 +44,30 @@ import {
   isOpenRouterDeepSeekHelloTerminalVariance,
   numberedPlanStepCount,
   providerSessionContinuityFailures,
+  reasoningProjectionFailures,
 } from "./run-observations.js";
 import { runnerE2EWebServerCommand } from "./web-server-command.js";
 
 const cleanupDirectories: string[] = [];
+
+it("rejects reasoning mislabeled as assistant text without echoing private content", () => {
+  const reasoning = {
+    eventType: "item.delta",
+    payload: { prpEvent: {
+      eventType: "item.delta",
+      payload: { kind: "agentMessage", text: "PRIVATE_THOUGHT", update: { kind: "reasoning" } },
+    } },
+  };
+  expect(reasoningProjectionFailures([reasoning])).toEqual([
+    "provider reasoning was projected as assistant text in 1 durable events",
+  ]);
+  const correctlyTyped = structuredClone(reasoning);
+  correctlyTyped.payload.prpEvent.payload.kind = "reasoning";
+  expect(reasoningProjectionFailures([correctlyTyped])).toEqual([]);
+  const assistant = structuredClone(reasoning);
+  assistant.payload.prpEvent.payload.update.kind = "agentMessage";
+  expect(reasoningProjectionFailures([assistant])).toEqual([]);
+});
 afterEach(async () => {
   vi.unstubAllEnvs();
   vi.unstubAllGlobals();