test: require recalled chat context after server restart

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

tests/runner-e2e/README.md

@@ -168,6 +168,10 @@ lifecycle/performance events are insufficient. The startup case must stop after
 a process launch request but before a provider turn starts. Missing the boundary
 fails the case instead of silently testing another phase.
 
+Restart continuity requires the agent to recall a phrase after the server
+restarts. The final prompt does not reveal that phrase. A generic successful
+reply after restart cannot pass this check.
+
 The committed-send case withholds the browser's comment acknowledgement until
 the agent has saved one backlog task, drops that response, restarts Paperclip,
 and replays the exact public request with the original client request ID. It

tests/runner-e2e/catalog.ts

@@ -948,7 +948,7 @@ export const runnerSuites: readonly RunnerSuiteFixture[] = [
     profiles: runnerProfiles.filter(profile => ["legacy-codex", "legacy-claude", "runner-codex", "runner-acpx-claude"].includes(profile.id)).map(defaultPermissionProfile),
     environments: [localEnvironment], tasks: chatTasks, expectedMatrixSize: 28,
     excludedExecutionIds: ["legacy-codex", "legacy-claude"].flatMap(profile => ["reassign-task", "create-backlog"].map(task => `agent-chat.${profile}.local.${task}`)),
-    definitionMetadata: { version: 5, resetRunsCountedSeparately: true, permissions: "production-defaults", stopBoundary: "provider-turn-started" },
+    definitionMetadata: { version: 6, resetRunsCountedSeparately: true, permissions: "production-defaults", stopBoundary: "provider-turn-started", restartMemory: "required-after-restart" },
   },
   {
     id: "agent-chat-hardening", label: "Agent Chat Recovery and Coordination", manualOnly: true,
@@ -959,7 +959,7 @@ export const runnerSuites: readonly RunnerSuiteFixture[] = [
     environments: [localEnvironment, daytonaWarmEnvironment], tasks: chatHardeningTasks, expectedMatrixSize: 18,
     excludedExecutionIds: ["runner-codex", "runner-acpx-claude"].flatMap(profile =>
       ["stop-startup-new-resume", "hire-delegate-reuse", "blocked-status-review"].map(task => `agent-chat-hardening.${profile}.daytona.${task}`)),
-    definitionMetadata: { version: 1, permissions: "production-defaults", instructions: "production", grading: "durable-state-and-source-evidence", scheduling: "explicit-only" },
+    definitionMetadata: { version: 2, permissions: "production-defaults", instructions: "production", grading: "durable-state-and-source-evidence", scheduling: "explicit-only", restartMemory: "required-after-restart" },
   },
   ...(process.env.PAPERCLIP_RUNNER_E2E_CONNECTION_REVIEWS === "1" ? [connectionReviewSuite] : []),
   {

tests/runner-e2e/chat-flow.ts

@@ -8,7 +8,7 @@ import type { LiveFixtureValues } from "./live-fixtures.js";
 import type { MatrixExecution } from "./types.js";
 import { isBlockedUnstartedWake } from "./non-execution-wake.js";
 import { chatMarker } from "./chat-cases.js";
-import { assertChatStartupStopped, isChatStopReady, runChatHardeningFlow } from "./chat-hardening.js";
+import { assertChatRememberedAfterRestart, assertChatStartupStopped, isChatStopReady, runChatHardeningFlow } from "./chat-hardening.js";
 
 // Public API observations only: this driver never fabricates provider results or writes DB state.
 export interface ChatIssue {
@@ -409,9 +409,14 @@ export async function runChatFlow(input: ChatFlowInput) {
         await idle(2);
         expect(runs).toHaveLength(count);
         await turn(
-          `We are done discussing it. Reply with ${marker} only; no further work.`,
+          `What phrase did I ask you to remember earlier? Reply with that remembered phrase followed by ${marker}; no further work.`,
           3,
         );
+        assertChatRememberedAfterRestart(
+          (await comments()).filter((comment) => comment.authorAgentId).at(-1)?.body ?? "",
+          secret,
+          marker,
+        );
       } else {
         let cancelledId: string | undefined;
         if (stopCase) {

tests/runner-e2e/chat-hardening.test.ts

@@ -1,5 +1,5 @@
 import { describe, expect, it } from "vitest";
-import { assertChatHire, assertChatSourceReview, assertChatStartupStopped, assertCommittedSendRetry, assertGroundedChatStatus, isChatStopReady } from "./chat-hardening.js";
+import { assertChatHire, assertChatRememberedAfterRestart, assertChatSourceReview, assertChatStartupStopped, assertCommittedSendRetry, assertGroundedChatStatus, isChatStopReady } from "./chat-hardening.js";
 import { runnerMatrix } from "./catalog.js";
 import { buildRunnerE2EProcessEnvironment } from "./harness-env.js";
 
@@ -10,6 +10,11 @@ const run = { id: "run", companyId: "company", agentId: "hire", status: "succeed
 const hiring = { agents: [{ id: "lead", name: "Lead", adapterConfig: { model: "model" } }, hire], leadId: "lead", hireName: "Morgan", hiredId: "hire", connectionId: "account", binding, taskIds: ["task"], tasks: [task], runs: [run] };
 
 describe("agent chat hardening oracles", () => {
+  it("requires remembered context after restart, not just a new reply", () => {
+    expect(() => assertChatRememberedAfterRestart("OLDCONTEXT123 CHAT123", "OLDCONTEXT123", "CHAT123")).not.toThrow();
+    expect(() => assertChatRememberedAfterRestart("CHAT123", "OLDCONTEXT123", "CHAT123")).toThrow();
+    expect(() => assertChatRememberedAfterRestart("OLDCONTEXT999 CHAT123", "OLDCONTEXT123", "CHAT123")).toThrow();
+  });
   it("grades actual execution by the original hired identity and managed account", () => {
     expect(() => assertChatHire(hiring)).not.toThrow();
     for (const wrong of [

tests/runner-e2e/chat-hardening.ts

@@ -76,6 +76,11 @@ export function isChatStopReady(events: Array<{ eventType?: unknown }>, phase: "
   return phase === "active" ? active : !active && events.some(event => event.eventType === "native.process_start_requested");
 }
 
+export function assertChatRememberedAfterRestart(reply: string, remembered: string, marker: string) {
+  expect(reply).toContain(remembered);
+  expect(reply).toContain(marker);
+}
+
 export function assertChatStartupStopped(run: ChatRun, events: Array<{ eventType?: unknown; createdAt?: string }>) {
   const cancellation = run.resultJson?.nativeCancellation as Record<string, unknown> | undefined;
   const started = events.filter(event => ["turn.started", "turn.accepted"].includes(String(event.eventType)));