feat(FN-3410): finalize planning verification and launcher updates

- Rework planning subtask descriptions and update planning logic/tests for clearer execution guidance
- Add launcher/update notice handling in cli-alias with package-config coverage and release script updates
- Refine agent detail and agents list UI/CSS for compressed mobile header behavior with updated dashboard tests
- Add changesets and related docs/readme touch-ups for release tracking
- Ref: runfusion/fusion#33

docs/task-management.md

@@ -22,6 +22,7 @@ Use the 💡 button to open planning mode:
 - AI reasoning (thinking output) is preserved and visible throughout the session — expand the reasoning toggle to review the model's analysis before answering each question or accepting the summary
 - Produces summary + key deliverables
 - Create one task or **Break into Tasks** (multi-task generation with dependencies)
+- Break-into-tasks descriptions are structured with subtask-specific guidance first, then a separate larger-plan context section (plus `## Planning Interview Context` when interview history exists)
 - Sessions persist when the modal is closed — resume from the sidebar list at any time; reasoning context is restored automatically
 
 ### 3) Todo item → Plan Mode

packages/dashboard/README.md

@@ -21,6 +21,7 @@ AI-guided interactive planning for creating well-specified tasks from high-level
    - Suggested dependencies from existing tasks
    - Key deliverables checklist
 5. Create the task directly from the summary
+6. Or use **Break into Tasks** to generate multiple subtasks where each description starts with subtask-specific implementation guidance, followed by a separate larger-plan context section (including planning interview context when available)
 
 **Features**:
 - **Rate Limiting**: Maximum 5 planning sessions per hour per IP

packages/dashboard/src/__tests__/planning.test.ts

@@ -1974,13 +1974,18 @@ describe("planning module", () => {
       });
     });
 
-    it("appends planning interview context to subtask descriptions when history exists", async () => {
+    it("generates deliverable subtasks with distinct lead guidance plus separate plan context", async () => {
       const mockIp = getUniqueIp();
       const sessionId = await createCompletedSession(mockIp, "Build auth system with context");
 
       const result = generateSubtasksFromPlanning(sessionId);
 
-      expect(result.length).toBeGreaterThan(0);
+      expect(result.length).toBe(3);
+      expect(result[0]?.description).toContain('Implement "Implementation" as this subtask\'s primary outcome.');
+      expect(result[1]?.description).toContain('Implement "Tests" as this subtask\'s primary outcome.');
+      expect(result[2]?.description).toContain('Implement "Documentation" as this subtask\'s primary outcome.');
+
+      expect(result[0]?.description).toContain("## Larger Plan Context");
       expect(result[0]?.description).toContain("## Planning Interview Context");
       expect(result[0]?.description).toContain("**Q: What is the scope of this plan?**");
       expect(result[0]?.description).toContain("A: Medium");
@@ -1990,7 +1995,7 @@ describe("planning module", () => {
       expect(result[0]?.description).toContain("A: Yes");
     });
 
-    it("keeps subtask descriptions unchanged when history is empty", async () => {
+    it("keeps larger-plan context section when history is empty", async () => {
       const mockIp = getUniqueIp();
       const sessionId = await createCompletedSession(mockIp, "Build auth without context");
 
@@ -2005,7 +2010,9 @@ describe("planning module", () => {
       const result = generateSubtasksFromPlanning(sessionId);
       expect(result.length).toBeGreaterThan(0);
       for (const subtask of result) {
-        expect(subtask.description).toBe(session.summary.description);
+        expect(subtask.description).toContain("## Larger Plan Context");
+        expect(subtask.description).toContain(session.summary.description);
+        expect(subtask.description).not.toContain("## Planning Interview Context");
       }
     });
 
@@ -2049,6 +2056,10 @@ describe("planning module", () => {
         suggestedSize: "S",
         dependsOn: ["subtask-2"],
       });
+      expect(result[0]?.description).toContain("Define the implementation approach for the plan");
+      expect(result[1]?.description).toContain("Implement the core code changes described by the plan");
+      expect(result[2]?.description).toContain("Verify the implementation end-to-end");
+      expect(result[0]?.description).toContain("## Larger Plan Context");
     });
 
     it("assigns correct sizes based on deliverable position", async () => {

packages/dashboard/src/planning.ts

@@ -2143,6 +2143,23 @@ export function getSummary(sessionId: string): PlanningSummary | undefined {
  * @param sessionId - The planning session ID
  * @returns Array of SubtaskItem with titles derived from keyDeliverables, or fallback
  */
+function buildPlanningSubtaskDescription(input: {
+  taskGuidance: string;
+  summaryDescription: string;
+  qaSection: string;
+}): string {
+  const contextSections = [
+    "## Larger Plan Context",
+    input.summaryDescription,
+  ];
+
+  if (input.qaSection) {
+    contextSections.push(input.qaSection);
+  }
+
+  return `${input.taskGuidance}\n\n${contextSections.join("\n\n")}`;
+}
+
 export function generateSubtasksFromPlanning(sessionId: string): SubtaskItem[] {
   const session = sessions.get(sessionId);
   if (!session) return [];
@@ -2150,9 +2167,6 @@ export function generateSubtasksFromPlanning(sessionId: string): SubtaskItem[] {
 
   const { summary } = session;
   const qaSection = formatInterviewQA(session.history);
-  const descriptionWithContext = qaSection
-    ? `${summary.description}\n\n${qaSection}`
-    : summary.description;
 
   // If key deliverables exist, create one subtask per deliverable
   if (summary.keyDeliverables.length > 0) {
@@ -2162,7 +2176,11 @@ export function generateSubtasksFromPlanning(sessionId: string): SubtaskItem[] {
       return {
         id,
         title: deliverable,
-        description: descriptionWithContext,
+        description: buildPlanningSubtaskDescription({
+          taskGuidance: `Implement "${deliverable}" as this subtask's primary outcome. Focus only on the concrete changes needed to deliver this item.`,
+          summaryDescription: summary.description,
+          qaSection,
+        }),
         suggestedSize: index === 0 ? "S" as const : index === summary.keyDeliverables.length - 1 ? "S" as const : "M" as const,
         dependsOn,
       };
@@ -2174,21 +2192,33 @@ export function generateSubtasksFromPlanning(sessionId: string): SubtaskItem[] {
     {
       id: "subtask-1",
       title: "Define implementation approach",
-      description: descriptionWithContext,
+      description: buildPlanningSubtaskDescription({
+        taskGuidance: "Define the implementation approach for the plan, including architecture and sequencing decisions needed before coding.",
+        summaryDescription: summary.description,
+        qaSection,
+      }),
       suggestedSize: "S" as const,
       dependsOn: [],
     },
     {
       id: "subtask-2",
       title: "Implement core changes",
-      description: descriptionWithContext,
+      description: buildPlanningSubtaskDescription({
+        taskGuidance: "Implement the core code changes described by the plan, using the agreed approach from the prior subtask.",
+        summaryDescription: summary.description,
+        qaSection,
+      }),
       suggestedSize: "M" as const,
       dependsOn: ["subtask-1"],
     },
     {
       id: "subtask-3",
       title: "Verify and polish",
-      description: descriptionWithContext,
+      description: buildPlanningSubtaskDescription({
+        taskGuidance: "Verify the implementation end-to-end, then polish quality items like tests, docs, and edge-case handling.",
+        summaryDescription: summary.description,
+        qaSection,
+      }),
       suggestedSize: "S" as const,
       dependsOn: ["subtask-2"],
     },