Scope warm acceptance instructions to the active runner protocol

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

tests/runner-e2e/catalog.test.ts

@@ -12,6 +12,7 @@ import {
   runnerSuites,
   runnerTasks,
   daytonaWarmContinuityTask,
+  warmPromptForGeneration,
   daytonaWarmEnvironment,
   isImmutableDaytonaImage,
   suiteDefinitionHash,
@@ -25,6 +26,20 @@ import {
 } from "./selectors.js";
 
 describe("runner E2E catalog", () => {
+  it("gives warm runners only their own completion protocol on all three turns", () => {
+    const prompts = [daytonaWarmContinuityTask.buildPrompt("nonce"), ...daytonaWarmContinuityTask.buildFollowupMessages!("nonce")];
+    for (const prompt of prompts) {
+      const native = warmPromptForGeneration(prompt, "native");
+      expect(native).toContain("paperclip_finish");
+      expect(native).not.toContain("PATCH /api/issues");
+      expect(native).not.toContain("In a legacy runner,");
+      const legacy = warmPromptForGeneration(prompt, "legacy");
+      expect(legacy).not.toContain("paperclip_finish");
+      expect(legacy).toContain("issue");
+      expect(legacy).toContain("Do not recreate, truncate, reorder, or duplicate prior lines.");
+    }
+  });
+
   it("defines sixteen local connection-review journeys without expanding the default matrix", () => {
     expect(connectionReviewSuite.expectedMatrixSize).toBe(16);
     expect(new Set(connectionReviewSuite.profiles.map(profile => profile.id))).toEqual(new Set(["runner-codex", "runner-acpx-claude", "legacy-codex", "legacy-claude"]));

tests/runner-e2e/catalog.ts

@@ -799,6 +799,14 @@ const localEnvironment = runnerEnvironments.find(
   (environment) => environment.id === "local",
 )!;
 
+export function warmPromptForGeneration(prompt: string, generation: RunnerProfileFixture["generation"]): string {
+  const other = generation === "native" ? "legacy" : "native";
+  return [
+    `You are running in a ${generation} runner. Follow the ${generation} completion instructions below.`,
+    ...prompt.split("\n").filter((line) => !line.startsWith(`In a ${other} runner,`)),
+  ].join("\n");
+}
+
 function warmTurnMarker(turn: 1 | 2 | 3, nonce: string) {
   return `PAPERCLIP_E2E_WARM_T${turn}_${nonce}`;
 }

tests/runner-e2e/runner.spec.ts

@@ -11,7 +11,7 @@ import path from "node:path";
 import { expect, test, type Page } from "@playwright/test";
 import { RunnerApi, pollUntil } from "./api.js";
 import { buildRuntimeUsage, summarizeExecutionBilling } from "./billing.js";
-import { runnerExecutionById } from "./catalog.js";
+import { runnerExecutionById, warmPromptForGeneration } from "./catalog.js";
 import { classifyFailure } from "./failure-classifier.js";
 import { runnerE2EServerControlPaths } from "./harness-env.js";
 import { setupConnectionReview } from "./connection-reviews.js";
@@ -525,7 +525,10 @@ for (const execution of executions) {
     const nonce = `${randomBytes(6).toString("hex")}-${attempt}`;
     const marker = execution.task.buildVisibleMarker(nonce);
     const title = execution.task.buildTitle(nonce);
-    const prompt = execution.task.buildPrompt(nonce);
+    const rawPrompt = execution.task.buildPrompt(nonce);
+    const prompt = execution.task.flow === "warm_three_turn"
+      ? warmPromptForGeneration(rawPrompt, execution.profile.generation)
+      : rawPrompt;
     const credentials = credentialValues();
     const secrets = normalizedSecrets(Object.values(credentials));
     const api = new RunnerApi(request);
@@ -1517,7 +1520,7 @@ for (const execution of executions) {
             `warm-turn-${completedTurn}.png`,
           );
           turnSubmissionTimesMs.push(
-            await submitTaskRevision(page, followups[completedTurn - 1]),
+            await submitTaskRevision(page, warmPromptForGeneration(followups[completedTurn - 1], execution.profile.generation)),
           );
         }
         warmLifecycleEvidence = { turns: turnEvidence };