FN-8646: fix dashboard task enrichment validation

Repair deterministic dashboard test regressions and reject whitespace-only enrichment targets.

- Normalize task IDs before enabling and submitting research enrichment.
- Update research, task-detail CSS, and chat visibility test fixtures.
- Add a patch changeset for the dashboard behavior fix.

Files changed:
 .changeset/fn-8646-dashboard-test-repairs.md         |  7 +++++++
 .../app/components/ResearchTaskActionModal.tsx       | 12 ++++++++++--
 .../app/components/__tests__/ResearchView.test.tsx   | 20 ++++++++++++++++----
 .../components/__tests__/TaskDetailModal.css.test.ts | 11 +++++++++--
 .../dashboard/app/hooks/__tests__/useChat.test.ts    |  9 ++++++++-
 5 files changed, 50 insertions(+), 9 deletions(-)

Fusion-Task-Id: FN-8646
Fusion-Task-Lineage: 033f4fa4-b096-4129-b82b-8fa7078dd6b2
Co-authored-by: Fusion (runfusion.ai) <noreply@runfusion.ai>

.changeset/fn-8646-dashboard-test-repairs.md

@@ -0,0 +1,7 @@
+---
+"@runfusion/fusion": patch
+---
+
+summary: Prevent blank task IDs from enabling Research finding enrichment.
+category: fix
+dev: Trims the target task ID before enabling and submitting the enrich action.

packages/dashboard/app/components/ResearchTaskActionModal.tsx

@@ -133,6 +133,14 @@ export function ResearchTaskActionModal({ open, mode, run, finding, projectId, o
     };
   }, [open, mode, projectId, isArchivedColumn]);
 
+  /*
+  FNXC:ResearchEnrich 2026-08-01-02:35:
+  Enriching requires a real task identifier. Whitespace made the action appear available and sent
+  a non-actionable identifier downstream, so validate the normalized value and submit that same
+  value rather than preserving an input-only representation.
+  */
+  const normalizedTaskId = taskId.trim();
+
   if (!open) return null;
 
   return (
@@ -195,11 +203,11 @@ export function ResearchTaskActionModal({ open, mode, run, finding, projectId, o
           <button
             className="btn btn-primary"
             type="button"
-            disabled={saving || (mode === "enrich" && !taskId)}
+            disabled={saving || (mode === "enrich" && !normalizedTaskId)}
             onClick={() => {
               setSaving(true);
               void onConfirm({
-                taskId: mode === "enrich" ? taskId : undefined,
+                taskId: mode === "enrich" ? normalizedTaskId : undefined,
                 title: mode === "create" ? title.trim() : undefined,
                 description: mode === "create" ? description.trim() : undefined,
                 priority: mode === "create" ? priority : undefined,

packages/dashboard/app/components/__tests__/ResearchView.test.tsx

@@ -1,5 +1,6 @@
 import { describe, it, expect, vi, beforeEach } from "vitest";
 import { fireEvent, render, screen, waitFor, within } from "@testing-library/react";
+import userEvent from "@testing-library/user-event";
 import { loadAllAppCss, loadAllAppCssBaseOnly } from "../../test/cssFixture";
 import { Header } from "../Header";
 import { ResearchView } from "../ResearchView";
@@ -291,7 +292,14 @@ describe("ResearchView", () => {
     });
   });
 
+  /*
+  FNXC:ResearchEnrich 2026-08-01-02:35:
+  The enrich control is a real editable combobox, not a broken product path. Drive it with
+  per-character user input so this regression test validates the same state updates as a user;
+  synthetic change events previously masked that interaction contract.
+  */
   it("triggers enrich-task action from finding modal", async () => {
+    const user = userEvent.setup();
     const attachRunToTask = vi.fn().mockResolvedValue({});
     mockUseResearch.mockReturnValue({
       ...baseHookValue,
@@ -314,8 +322,8 @@ describe("ResearchView", () => {
 
     const enrichDialog = await screen.findByRole("dialog");
     const targetInput = within(enrichDialog).getByRole("combobox", { name: "Target task" });
-    fireEvent.change(targetInput, { target: { value: "FN-1" } });
+    await user.type(targetInput, "FN-1");
-    fireEvent.click(within(enrichDialog).getByRole("button", { name: "Enrich Task" }));
+    await user.click(within(enrichDialog).getByRole("button", { name: "Enrich Task" }));
 
     await waitFor(() => {
       expect(attachRunToTask).toHaveBeenCalledWith("RR-1", "FN-1", "finding-1", false);
@@ -533,7 +541,8 @@ describe("ResearchView", () => {
     });
   });
 
-  it("keeps enrich action disabled until a task id is provided", async () => {
+  it("keeps enrich action disabled for empty and whitespace-only task ids, then enables for a typed id", async () => {
+    const user = userEvent.setup();
     mockUseResearch.mockReturnValue({
       ...baseHookValue,
       runs: [{ id: "RR-1", title: "t", query: "q", status: "completed" }],
@@ -557,7 +566,10 @@ describe("ResearchView", () => {
     expect(enrichButton).toBeDisabled();
 
     const targetInput = within(dialog).getByRole("combobox", { name: "Target task" });
-    fireEvent.change(targetInput, { target: { value: "FN-1" } });
+    await user.type(targetInput, "   ");
+    expect(enrichButton).toBeDisabled();
+    await user.clear(targetInput);
+    await user.type(targetInput, "FN-1");
     await waitFor(() => expect(enrichButton).not.toBeDisabled());
   });
 

packages/dashboard/app/components/__tests__/TaskDetailModal.css.test.ts

@@ -28,13 +28,20 @@ describe("TaskDetailModal CSS contract", () => {
     expect(css).toMatch(/\.detail-tab\s*\{[^}]*flex-shrink\s*:\s*0\s*;/);
   });
 
+  /*
+  FNXC:TaskDetailPadding 2026-08-01-03:20:
+  FN-8634 moved the canonical tab inset from the scrolling `.detail-body` to its
+  `.detail-body-content` child so scrollbar tracks remain outside the content inset.
+  FN-7408 must assert that visual contract at the owning element while chat and
+  planner variants remain forbidden from introducing competing outer padding.
+  */
   it("FN-7408 keeps task-detail tab body padding canonical across Activity, planner Chat, and Plan surfaces", async () => {
     const css = await loadAllAppCssBaseOnly();
-    const detailBodyBlock = getCssRuleBlock(css, ".detail-body");
+    const detailBodyContentBlock = getCssRuleBlock(css, ".detail-body-content");
     const rawBodyBlock = getCssRuleBlock(css, ".detail-body--agent-log");
     const planBlock = getCssRuleBlock(css, ".detail-section--plan-prompt");
 
-    expect(detailBodyBlock).toContain("padding: calc(var(--space-lg) + var(--space-xs));");
+    expect(detailBodyContentBlock).toContain("padding: calc(var(--space-lg) + var(--space-xs));");
     expectNoOuterPaddingOverride(css, ".detail-body--chat");
     expectNoOuterPaddingOverride(css, ".detail-body--planner-chat");
     expectNoOuterPaddingOverride(css, ".task-detail-content--chat-expanded .detail-body--chat");

packages/dashboard/app/hooks/__tests__/useChat.test.ts

@@ -1827,7 +1827,14 @@ describe("useChat", () => {
       },
     };
     mockFetchChatSessions.mockResolvedValueOnce({ sessions: [session] });
-    mockFetchChatSession.mockResolvedValueOnce({ session: generatingSession });
+    /*
+    FNXC:ChatVisibilityResume 2026-08-01-02:35:
+    The visible-edge reconciliation now performs its own authoritative session read after
+    selection. Keep both reads generating: a one-shot mock made the second read fall back to an
+    unrelated idle fixture, which falsely modeled a completed user-visible generation and hid the
+    prior thread this test is meant to protect.
+    */
+    mockFetchChatSession.mockResolvedValue({ session: generatingSession });
     mockFetchChatMessages
       .mockResolvedValueOnce({ messages: [] })
       .mockResolvedValueOnce({ messages: priorThreadNewestFirst });