fix(managed_agents): address remaining memory cookbook review items

- Add wait_for_idle_status calls before session archive in cleanup
- Break out of run_turn on any status_idle, not just end_turn
- Guard event.input against None in tool_use handler
- Move seeding-order warning to a markdown Note callout
- Add %pip install cell for minimum SDK version
- Link to sibling CMA notebooks from the summary

Author: gaganb-ant

managed_agents/CMA_remember_user_preferences.ipynb

@@ -52,6 +52,17 @@
     "## Set up the client"
    ]
   },
+  {
+   "cell_type": "code",
+   "id": "da6a39d52f3e4b0a",
+   "metadata": {},
+   "execution_count": null,
+   "outputs": [],
+   "source": [
+    "%%capture\n",
+    "%pip install -q \"anthropic>=0.91.0\""
+   ]
+  },
   {
    "cell_type": "code",
    "id": "3f060791d5fa56fd8657f80c37f2d982",
@@ -111,13 +122,18 @@
     "            elif event.type == \"agent.tool_use\":\n",
     "                # Surface reads and writes to the memory mount so you can see\n",
     "                # the agent checking and updating its memory.\n",
-    "                target = event.input.get(\"file_path\") or event.input.get(\"command\", \"\")\n",
+    "                inp = event.input or {}\n",
+    "                target = inp.get(\"file_path\") or inp.get(\"command\", \"\")\n",
     "                if \"/mnt/memory/\" in str(target):\n",
     "                    print(f\"  [memory] {event.name}: {target}\")\n",
     "\n",
     "            elif event.type == \"session.status_idle\":\n",
     "                if event.stop_reason and event.stop_reason.type == \"end_turn\":\n",
     "                    break\n",
+    "                # This notebook only sends user.message events, so end_turn is the\n",
+    "                # only idle reason we expect. Break on anything else as well so an\n",
+    "                # unexpected stop_reason (such as requires_action) cannot hang the loop.\n",
+    "                break\n",
     "\n",
     "            elif event.type == \"session.status_terminated\":\n",
     "                break\n",
@@ -483,7 +499,9 @@
     "\n",
     "### Seed a store from your existing data\n",
     "\n",
-    "If you already know things about a customer from their account profile or purchase history, you can write them into the store before the first session so the agent starts informed. In a real application you would run this seeding step before any sessions are created; it appears here, after the demo, only so the main learn-then-recall flow above stays focused."
+    "If you already know things about a customer from their account profile or purchase history, you can write them into the store before the first session so the agent starts informed. In a real application you would run this seeding step before any sessions are created; it appears here, after the demo, only so the main learn-then-recall flow above stays focused.\n",
+    "\n",
+    "> **Note:** Run this cell before the cleanup cell at the end of the notebook, since `store` is deleted there."
    ]
   },
   {
@@ -501,7 +519,6 @@
     }
    ],
    "source": [
-    "# Run before the cleanup cell below; `store` is deleted there.\n",
     "seeded = client.beta.memory_stores.memories.create(\n",
     "    store.id,\n",
     "    path=\"/purchase-history.md\",\n",
@@ -571,6 +588,9 @@
    "execution_count": 10,
    "outputs": [],
    "source": [
+    "wait_for_idle_status(client, session_one.id)\n",
+    "wait_for_idle_status(client, session_two.id)\n",
+    "\n",
     "client.beta.sessions.archive(session_one.id)\n",
     "client.beta.sessions.archive(session_two.id)\n",
     "client.beta.memory_stores.delete(store.id)\n",
@@ -594,6 +614,11 @@
     "\n",
     "From here you can map your own user IDs to memory store IDs, seed stores from your existing customer data, and layer shared read-only stores on top for brand-wide knowledge.\n",
     "\n",
+    "### Other notebooks in this series\n",
+    "\n",
+    "- [`CMA_iterate_fix_failing_tests.ipynb`](CMA_iterate_fix_failing_tests.ipynb) \u2014 the entry-point notebook. Introduces agents, environments, sessions, file mounts, and the streaming event loop through a do-observe-fix loop on a failing test suite.\n",
+    "- [`CMA_operate_in_production.ipynb`](CMA_operate_in_production.ipynb) \u2014 production setup story: vault-backed MCP credentials, the `session.status_idled` webhook for HITL without long-lived connections, and the resource lifecycle CRUD verbs.\n",
+    "\n",
     "### Learn more\n",
     "\n",
     "- [Claude Managed Agents overview](https://docs.anthropic.com/en/docs/managed-agents/overview)\n",