[grade=B] fix(DC-163 rev2): strict merge semantics, contract-proving settings/test, valid enginecheck fixture

Judge r1 (C) blockers B1-B3 + polish both applied — see body of rev1 commit for feature scope.

urn:ump:l2tg66wmchkmd6fde4k3ojwc56zpxqh657spe37bu6dpvduswalq

dashcaddy-api/__tests__/routes/ai-intent-engine.test.js

@@ -180,6 +180,27 @@ describe('DC-163 classify engine selection', () => {
     expect(out.appId).toBe('uptime-kuma'); // NOT in the old regex list — catalog-wide
   });
 
+  test('LLM reroute: pattern deploy result + LLM unknown → action null, stale appId dropped', async () => {
+    // Judge r1 blocker #1: when the LLM overrides the pattern's intent,
+    // NO intent-specific field from the pattern may ride along.
+    const cm = fakeCredStore(JSON.stringify({ baseUrl: 'http://llm.test/v1', apiKey: '***', model: 'test-model' }));
+    const out = await intentMod.classify({
+      message: 'Deploy Plex', // pattern: deploy + appId plex
+      history: [],
+      deps: { credentialManager: cm, env: {}, fetchImpl: llmReplyFetch('{"intent":"rm -rf","appId":"plex","reply":"hmm"}'), log: silentLog, timeoutMs: 5000 },
+    });
+    expect(out.engine).toBe('llm');
+    expect(out.intent).toBe('unknown'); // illegal value coerced
+    // action must be strictly null (unknown has no action) — never the
+    // pattern's deploy action riding along
+    expect(out.action).toBeNull();
+    // stale pattern appId must NOT be carried under the new intent
+    expect(out.appId === undefined || out.appId === null).toBe(true);
+    const enriched = intentMod.enrich(out);
+    expect(enriched.response.message).toBeTruthy(); // rebuilt for unknown, no deployPlan
+    expect(enriched.deployPlan).toBeUndefined();
+  });
+
   test('LLM disagreement: pattern said recommend, LLM says deploy → deploy response rebuilt', async () => {
     const cm = fakeCredStore(JSON.stringify({ baseUrl: 'http://llm.test/v1', apiKey: LLM_TEST_KEY }));
     const out = await intentMod.classify({
@@ -352,6 +373,23 @@ describe('DC-163 ai HTTP routes', () => {
     } finally { server.close(); }
   });
 
+  test('POST /ai/settings/test — sample MUST classify as backup; coerced unknown = failure', async () => {
+    // Judge r1 blocker #2: accepting a coerced 'unknown' would let a model
+    // that ignores our prompt "pass" the test. Strict expectation now.
+    const cm = fakeCredStore(JSON.stringify({ baseUrl: 'http://llm.test/v1', apiKey: LLM_TEST_KEY, model: 'test-model' }));
+    const fetchImpl = llmReplyFetch('{"intent":"chitchat","appId":null,"reply":"lol sure"}'); // illegal -> coerces to unknown
+    const app = buildRouter({ credentialManager: cm, fetchImpl });
+    const { server, base } = await start(app);
+    try {
+      const res = await fetch(`${base}/ai/settings/test`, { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: '{}' });
+      const body = await res.json();
+      expect(res.status).toBe(200);
+      expect(body.ok).toBe(false);
+      expect(body.engine).toBe('llm-error');
+      expect(body.reason).toMatch(/expected 'backup', got 'unknown'/);
+    } finally { server.close(); }
+  });
+
   test('GET /ai/capabilities — engine reflects live configuration', async () => {
     const app = buildRouter();
     const { server, base } = await start(app);