feat(FN-4041): harden empty rawResponse handling in memory-insights

Hardened empty `rawResponse` handling in the memory insights module by adding defensive checks to prevent errors when the response is empty, accompanied by tests to cover those edge cases (FN-4041 Step 1).

Fusion-Task-Id: FN-4041

packages/core/src/__tests__/memory-insights.test.ts

@@ -838,6 +838,19 @@ describe("memory-insights run processing", () => {
       expect(result.summary).toContain("AI timeout");
     });
 
+    it("should treat undefined successful response as no output", async () => {
+      const result = await processInsightExtractionRun(tempDir, {
+        rawResponse: undefined,
+        stepSuccess: true,
+        runAt: new Date().toISOString(),
+      });
+
+      expect(result.insights).toHaveLength(0);
+      expect(result.summary).toBe("Step did not produce output");
+      expect(result.newInsightCount).toBe(0);
+      expect(result.duplicateCount).toBe(0);
+    });
+
     it("should preserve existing insights on failure", async () => {
       // Create existing insights
       const existingInsights = `# Memory Insights

packages/core/src/memory-insights.ts

@@ -230,7 +230,7 @@ interface MemoryAuditState {
 /** Input for processing an insight extraction run. */
 export interface ProcessRunInput {
   /** Raw AI response text from the insight extraction step. */
-  rawResponse: string;
+  rawResponse?: string;
   /** Whether the AI step itself succeeded. */
   stepSuccess: boolean;
   /** Timestamp of the run. */
@@ -1118,7 +1118,7 @@ export async function processInsightExtractionRun(
   let parseError: string | undefined;
 
   // Try to parse the AI response
-  if (stepSuccess && rawResponse.trim()) {
+  if (stepSuccess && rawResponse?.trim()) {
     try {
       parsedResult = parseInsightExtractionResponse(rawResponse);
     } catch (err) {

packages/engine/src/__tests__/triage.test.ts

@@ -1623,6 +1623,53 @@ describe("approved triage recovery", () => {
     );
   });
 
+  it("preserves imported GitHub issue titles during planning recovery", async () => {
+    await writeFile(
+      join(rootDir, ".fusion", "tasks", "FN-001", "PROMPT.md"),
+      "# Task: FN-001 - Different AI-generated planning title\n\n**Size:** M\n\n## Review Level: 2\n\nRecovered specification",
+    );
+
+    const store = createMockStore({
+      getSettings: vi.fn().mockResolvedValue({
+        maxConcurrent: 2,
+        maxWorktrees: 4,
+        pollIntervalMs: 10000,
+        groupOverlappingFiles: false,
+        autoMerge: true,
+        requirePlanApproval: false,
+      } as Settings),
+    });
+
+    const processor = new TriageProcessor(store, rootDir);
+    const recovered = await processor.recoverApprovedTask({
+      id: "FN-001",
+      description: "Imported from GitHub",
+      column: "triage",
+      status: "planning",
+      title: '"Cannot read properties of undefined (reading \'trim\')" when extracting insights',
+      sourceType: "github_import",
+      sourceIssue: {
+        provider: "github",
+        repository: "Runfusion/Fusion",
+        externalIssueId: "70",
+        issueNumber: 70,
+        url: "https://github.com/Runfusion/Fusion/issues/70",
+      },
+      dependencies: [],
+      steps: [],
+      currentStep: 0,
+      log: [{ timestamp: "2026-01-01T00:00:00.000Z", action: "Spec review: APPROVE" }],
+      createdAt: "2026-01-01T00:00:00.000Z",
+      updatedAt: "2026-01-01T00:02:00.000Z",
+    });
+
+    expect(recovered).toBe(true);
+    expect(store.updateTask).not.toHaveBeenCalledWith(
+      "FN-001",
+      expect.objectContaining({ title: "Different AI-generated planning title" }),
+    );
+  });
+
   it("clears status and error before moving approved tasks to todo", async () => {
     const store = createMockStore({
       getSettings: vi.fn().mockResolvedValue({

packages/engine/src/triage.ts

@@ -2064,12 +2064,13 @@ export class TriageProcessor {
     // ordering as defense in depth so a future change to the guard can't
     // resurrect the regression.
     const promptDeclaredTitle = extractPromptDeclaredTitle(written, task.id);
+    const shouldApplyPromptDeclaredTitle = shouldReplaceTaskTitleFromPrompt(task, promptDeclaredTitle);
 
     await this.store.updateTask(task.id, taskUpdates);
 
     if (settings.requirePlanApproval) {
       const approvalUpdates: Record<string, unknown> = { status: "awaiting-approval" };
-      if (promptDeclaredTitle) {
+      if (shouldApplyPromptDeclaredTitle && promptDeclaredTitle) {
         approvalUpdates.title = promptDeclaredTitle;
       }
       await this.store.updateTask(task.id, approvalUpdates);
@@ -2083,7 +2084,7 @@ export class TriageProcessor {
 
     await this.store.moveTask(task.id, "todo");
 
-    if (promptDeclaredTitle) {
+    if (shouldApplyPromptDeclaredTitle && promptDeclaredTitle) {
       await this.store.updateTask(task.id, { title: promptDeclaredTitle });
     }
 
@@ -2112,13 +2113,32 @@ function extractPromptDeclaredTitle(prompt: string, taskId: string): string | nu
   if (!title) return null;
 
   // Conservative guard: do not overwrite metadata with confirmation prose.
-  if (/^created\s+(?:task\s+)?(?:fn-\d+\b|\*\*\s*fn-\d+\s*\*\*)/i.test(title)) {
+  if (isMalformedTaskTitle(title)) {
     return null;
   }
 
   return title;
 }
 
+function isMalformedTaskTitle(title: string): boolean {
+  return /^created\s+(?:task\s+)?(?:fn-\d+\b|\*\*\s*fn-\d+\s*\*\*)/i.test(title.trim());
+}
+
+function shouldReplaceTaskTitleFromPrompt(task: Task, promptDeclaredTitle: string | null): boolean {
+  if (!promptDeclaredTitle) return false;
+
+  if (
+    task.sourceType === "github_import" &&
+    task.sourceIssue?.provider === "github" &&
+    task.title?.trim() &&
+    !isMalformedTaskTitle(task.title)
+  ) {
+    return false;
+  }
+
+  return true;
+}
+
 function hasLatestSpecReviewApproval(task: Task): boolean {
   for (let i = task.log.length - 1; i >= 0; i--) {
     const action = task.log[i]?.action ?? "";