test: PTY harness handles clipped reviews and bundled setup tabs; AUQ judge uses structured output; design-consultation carve declines optional outside voices

- ceo mode routing: a Submit review taller than the viewport, a setup tab
  bundled after the mode tab, and a clip through the mode question each hung
  or misread the run; the native answer is still verified after Submit.
- judgeRecommendation requests a 1-5 enum schema; a malformed Haiku reply had
  scored substance 0 for a 4/5 brief. Judge failures now propagate.
- carve section-loading for design-consultation declines the optional outside
  voices (a supported path) and treats DESIGN.md as the report; timeout unchanged.
The Step 0E handoff defect is not fixed (0/15 samples across four wordings,
none shipped) and is filed in TODOS.

test/llm-judge-frontier.test.ts

@@ -1,6 +1,7 @@
 import { afterEach, beforeEach, describe, expect, spyOn, test } from 'bun:test';
 import Anthropic from '@anthropic-ai/sdk';
-import { armJudge, callJudge, JudgeRefusalError } from './helpers/llm-judge';
+import { armJudge, callJudge, JudgeRefusalError, judgeRecommendation, RECOMMENDATION_JUDGE_SCHEMA } from './helpers/llm-judge';
+import { gradeAuqRecommendation } from './helpers/auq-sdk-capture';
 
 describe('frontier Claude judge compatibility', () => {
   let originalKey: string | undefined;
@@ -194,6 +195,35 @@ describe('frontier Claude judge compatibility', () => {
     }
   });
 
+  // Census 36641820398 slice 7: Haiku's free-form reply left inner quotes unescaped.
+  const CENSUS_UNESCAPED_REPLY = '```json\n{"reason_substance": 4, "reasoning": "The clause is concrete and option-specific ("at 1/10 every dimension has gaps and none are obviously safe to skip") but does not explicitly compare the chosen option (A) against alternatives B or C in the because-clause itself\u2014the comparison lives in the surrounding context, not in the reason block."}\n```';
+  const CENSUS_BRIEF = 'D1 \u2014 Review all 7 design dimensions?\nRecommendation: A because at 1/10 every dimension has gaps and none are obviously safe to skip.\nA) All 7 dimensions (recommended)';
+
+  test('recommendation judge requests structured output, so quoted evidence cannot break its JSON', async () => {
+    create.mockResolvedValue({ stop_reason: 'end_turn', content: [{ type: 'text',
+      text: JSON.stringify({ reason_substance: 4, reasoning: 'Concrete ("at 1/10 every dimension has gaps") but no comparison.' }) }] } as never);
+    const score = await judgeRecommendation(CENSUS_BRIEF);
+    expect(score.reason_substance).toBe(4);
+    expect(score.reason_text).toBe('at 1/10 every dimension has gaps and none are obviously safe to skip.');
+    expect(create.mock.calls[0][0].output_config).toEqual({ format: { type: 'json_schema', schema: RECOMMENDATION_JUDGE_SCHEMA } });
+    expect(create.mock.calls[0][0].model).toBe('claude-haiku-4-5-20251001');
+  });
+
+  test('the captured unescaped reply is unparseable free-form text and fails without a score', async () => {
+    create.mockResolvedValue({ stop_reason: 'end_turn', content: [{ type: 'text', text: CENSUS_UNESCAPED_REPLY }] } as never);
+    await expect(callJudge('score this', 'claude-haiku-4-5-20251001')).rejects.toThrow(SyntaxError);
+    create.mockClear();
+    create.mockResolvedValue({ stop_reason: 'end_turn', content: [{ type: 'text', text: CENSUS_UNESCAPED_REPLY }] } as never);
+    await expect(gradeAuqRecommendation(CENSUS_BRIEF)).rejects.toThrow(SyntaxError);
+    expect(create).toHaveBeenCalledTimes(1);
+  });
+
+  test('a weak structured score still fails the substance bar', async () => {
+    create.mockResolvedValue({ stop_reason: 'end_turn', content: [{ type: 'text',
+      text: JSON.stringify({ reason_substance: 1, reasoning: 'Boilerplate.' }) }] } as never);
+    expect((await gradeAuqRecommendation('Recommendation: A because it is better.')).substance).toBe(1);
+  });
+
   test('arm judge sends no unsupported temperature to Fable', async () => {
     create.mockResolvedValue({ content: [
       { type: 'thinking', thinking: '', signature: 'fixture' },