Tighten parity proof heuristics

extensions/qa-lab/src/agentic-parity-report.test.ts

@@ -130,6 +130,27 @@ describe("qa agentic parity report", () => {
     );
   });
 
+  it("ignores neutral Failed and Blocked headings in passing protocol reports", () => {
+    const summary: QaParitySuiteSummary = {
+      scenarios: [
+        {
+          name: "Source and docs discovery report",
+          status: "pass",
+          details: `Worked:
+- Read the seeded QA material.
+Failed:
+- None observed.
+Blocked:
+- No live provider evidence in this lane.
+Follow-up:
+- Re-run with a real provider if needed.`,
+        },
+      ],
+    };
+
+    expect(computeQaAgenticParityMetrics(summary).fakeSuccessCount).toBe(0);
+  });
+
   it("renders a readable markdown parity report", () => {
     const comparison = buildQaAgenticParityComparison({
       candidateLabel: "openai/gpt-5.4",

extensions/qa-lab/src/agentic-parity-report.ts

@@ -68,9 +68,11 @@ const SUSPICIOUS_PASS_PATTERNS = [
   /incomplete turn/i,
   /\btimed out\b/i,
   /\btimeout\b/i,
-  /\bblocked\b/i,
   /\berror\b/i,
-  /\bfailed\b/i,
+  /\bfailed to\b/i,
+  /\bcould not\b/i,
+  /\bunable to\b/i,
+  /did not continue/i,
 ] as const;
 
 function normalizeScenarioStatus(status: string | undefined): "pass" | "fail" | "skip" {

extensions/qa-lab/src/mock-openai-server.test.ts

@@ -228,7 +228,7 @@ describe("qa mock openai server", () => {
           },
           {
             type: "function_call_output",
-            output: "Replay safety: unsafe after write.\n",
+            output: "Successfully wrote 41 bytes to compaction-retry-summary.txt.",
           },
         ],
       }),

extensions/qa-lab/src/mock-openai-server.ts

@@ -453,7 +453,12 @@ function buildAssistantText(input: ResponsesInputItem[], body: Record<string, un
     return `Protocol note: Lobster Invaders built at lobster-invaders.html.`;
   }
   if (toolOutput && /compaction retry mutating tool check/i.test(prompt)) {
-    if (toolOutput.includes("Replay safety: unsafe after write.")) {
+    if (
+      toolOutput.includes("Replay safety: unsafe after write.") ||
+      /compaction-retry-summary\.txt/i.test(toolOutput) ||
+      /successfully (?:wrote|replaced)/i.test(toolOutput) ||
+      /\bwrote\b.*\bcompaction-retry-summary\.txt\b/i.test(toolOutput)
+    ) {
       return "Protocol note: replay unsafe after write.";
     }
     return "";