diff --git a/checkpoint.json b/checkpoint.json
index 89b04aa9b..4b25929b5 100644
--- a/checkpoint.json
+++ b/checkpoint.json
@@ -1,31 +1,38 @@
 {
-  "timestamp": "2026-03-19T13:28:22.243831Z",
-  "current_node": "setup",
+  "timestamp": "2026-03-19T13:28:31.323789Z",
+  "current_node": "work",
   "completed_nodes": [
     "start",
-    "setup"
+    "setup",
+    "work"
   ],
   "node_retries": {
     "start": 1,
-    "setup": 1
+    "setup": 1,
+    "work": 1
   },
   "context_values": {
     "graph.rankdir": "LR",
     "internal.run_id": "01KM34GS7MYBV3D0YN2V19YPXZ",
     "graph.goal": "Verify sandbox filesystem is shared across command and agent nodes",
-    "current.preamble": "Goal: Verify sandbox filesystem is shared across command and agent nodes\n",
+    "current.preamble": "Goal: Verify sandbox filesystem is shared across command and agent nodes\n\n## Completed stages\n- **setup**: success\n  - Script: `echo SCENARIO_FLAG_42 > /tmp/scenario_flag.txt`\n  - Stdout: (empty)\n  - Stderr: (empty)\n",
     "failure_signature": "",
     "thread.start.current_node": "setup",
     "command.output": "",
+    "thread.setup.current_node": "work",
+    "internal.retry_count.work": 1,
     "outcome": "success",
     "failure_class": "",
     "internal.fidelity": "compact",
-    "current_node": "setup",
-    "internal.thread_id": "start",
+    "current_node": "work",
+    "internal.thread_id": "setup",
     "internal.retry_count.start": 1,
+    "response.work": "Perfect! The file `/tmp/scenario_flag.txt` contains:\n\n```\nSCENARIO_FLAG_42\n```\n\nThis confirms that the sandbox filesystem is shared across nodes. The setup stage successfully created the file with the content `SCENARIO_FLAG_42`, and I was able to read it from the current context. This demonstrates that the filesystem is properly shared and accessible across different stages/nodes in the workflow.",
     "internal.node_visit_count": 1,
+    "last_response": "Perfect! The file `/tmp/scenario_flag.txt` contains:\n\n```\nSCENARIO_FLAG_42\n```\n\nThis confirms that the sandbox filesystem is shared across nodes. The setup stage successfully created the file with the",
     "command.stderr": "",
     "internal.work_dir": "/private/var/folders/nm/bd2dwzy52t1ckn3tklsfmy1c0000gn/T/.tmpQYUGzd/run/worktree",
+    "last_stage": "work",
     "internal.retry_count.setup": 1
   },
   "logs": [],
@@ -42,11 +49,30 @@
     "start": {
       "status": "success",
       "duration_ms": 0
+    },
+    "work": {
+      "status": "success",
+      "context_updates": {
+        "last_response": "Perfect! The file `/tmp/scenario_flag.txt` contains:\n\n```\nSCENARIO_FLAG_42\n```\n\nThis confirms that the sandbox filesystem is shared across nodes. The setup stage successfully created the file with the",
+        "last_stage": "work",
+        "response.work": "Perfect! The file `/tmp/scenario_flag.txt` contains:\n\n```\nSCENARIO_FLAG_42\n```\n\nThis confirms that the sandbox filesystem is shared across nodes. The setup stage successfully created the file with the content `SCENARIO_FLAG_42`, and I was able to read it from the current context. This demonstrates that the filesystem is properly shared and accessible across different stages/nodes in the workflow."
+      },
+      "notes": "Stage completed: work",
+      "usage": {
+        "model": "claude-haiku-4-5",
+        "input_tokens": 509,
+        "output_tokens": 168,
+        "cache_read_tokens": 4745,
+        "cache_write_tokens": 5150,
+        "cost": 0.0010792
+      },
+      "duration_ms": 6545
     }
   },
-  "next_node_id": "work",
+  "next_node_id": "verify",
   "node_visits": {
     "setup": 1,
+    "work": 1,
     "start": 1
   }
 }
\ No newline at end of file
diff --git a/nodes/work/prompt.md b/nodes/work/prompt.md
new file mode 100644
index 000000000..59af50d8f
--- /dev/null
+++ b/nodes/work/prompt.md
@@ -0,0 +1,10 @@
+Goal: Verify sandbox filesystem is shared across command and agent nodes
+
+## Completed stages
+- **setup**: success
+  - Script: `echo SCENARIO_FLAG_42 > /tmp/scenario_flag.txt`
+  - Stdout: (empty)
+  - Stderr: (empty)
+
+
+Read the file /tmp/scenario_flag.txt and tell me what it contains.
\ No newline at end of file
diff --git a/nodes/work/provider_used.json b/nodes/work/provider_used.json
new file mode 100644
index 000000000..17e74f4d1
--- /dev/null
+++ b/nodes/work/provider_used.json
@@ -0,0 +1,5 @@
+{
+  "mode": "agent",
+  "provider": "anthropic",
+  "model": "claude-haiku-4-5"
+}
\ No newline at end of file
diff --git a/nodes/work/response.md b/nodes/work/response.md
new file mode 100644
index 000000000..a00bba732
--- /dev/null
+++ b/nodes/work/response.md
@@ -0,0 +1,7 @@
+Perfect! The file `/tmp/scenario_flag.txt` contains:
+
+```
+SCENARIO_FLAG_42
+```
+
+This confirms that the sandbox filesystem is shared across nodes. The setup stage successfully created the file with the content `SCENARIO_FLAG_42`, and I was able to read it from the current context. This demonstrates that the filesystem is properly shared and accessible across different stages/nodes in the workflow.
\ No newline at end of file
diff --git a/nodes/work/status.json b/nodes/work/status.json
new file mode 100644
index 000000000..6cfdc9fff
--- /dev/null
+++ b/nodes/work/status.json
@@ -0,0 +1,6 @@
+{
+  "status": "success",
+  "notes": "Stage completed: work",
+  "failure_reason": null,
+  "timestamp": "2026-03-19T13:28:31.323396+00:00"
+}
\ No newline at end of file