diff --git a/app-instance/backend/beaver/engine/loop.py b/app-instance/backend/beaver/engine/loop.py
index 35ce61d..235e0fc 100644
--- a/app-instance/backend/beaver/engine/loop.py
+++ b/app-instance/backend/beaver/engine/loop.py
@@ -644,29 +644,32 @@ class AgentLoop:
                 }
                 if thinking_enabled is not None:
                     chat_kwargs["thinking_enabled"] = thinking_enabled
+                message_char_length = len(json.dumps(messages, ensure_ascii=False, default=str))
+                tool_schema_char_length = len(json.dumps(tool_schemas, ensure_ascii=False, default=str))
+                tool_names = [
+                    str(tool.get("function", {}).get("name") or tool.get("name") or "tool")
+                    for tool in (tool_schemas or [])
+                    if isinstance(tool, dict)
+                ]
+                snapshot_payload = {
+                    "iteration": iterations,
+                    "provider_name": final_provider_name,
+                    "model": final_model,
+                    "message_count": len(messages),
+                    "tool_names": tool_names,
+                    "message_char_length": message_char_length,
+                    "tool_schema_char_length": tool_schema_char_length,
+                    "max_tokens": resolved_max_tokens,
+                    "temperature": resolved_temperature,
+                    "thinking_enabled": thinking_enabled,
+                }
                 session_manager.append_message(
                     resolved_session_id,
                     run_id=resolved_run_id,
                     role="system",
                     event_type="llm_request_snapshotted",
-                    event_payload={
-                        "iteration": iterations,
-                        "provider_name": final_provider_name,
-                        "model": final_model,
-                        "messages": messages,
-                        "tools": tool_schemas,
-                        "max_tokens": resolved_max_tokens,
-                        "temperature": resolved_temperature,
-                        "thinking_enabled": thinking_enabled,
-                    },
-                    content=json.dumps(
-                        {
-                            "messages": messages,
-                            "tools": tool_schemas,
-                        },
-                        ensure_ascii=False,
-                        default=str,
-                    ),
+                    event_payload=snapshot_payload,
+                    content=json.dumps(snapshot_payload, ensure_ascii=False, default=str),
                     context_visible=False,
                     source=source,
                     title=title,
diff --git a/app-instance/backend/tests/unit/test_phase5_skills_runtime.py b/app-instance/backend/tests/unit/test_phase5_skills_runtime.py
index 29e810a..08a6242 100644
--- a/app-instance/backend/tests/unit/test_phase5_skills_runtime.py
+++ b/app-instance/backend/tests/unit/test_phase5_skills_runtime.py
@@ -633,3 +633,23 @@ def test_agent_loop_records_max_tool_iterations_as_failed_skill_effect(tmp_path:
     effect_records = loaded.run_memory_store.list_skill_effects("docker-debug", version="v0007")
     assert effect_records[-1].run_id == result.run_id
     assert effect_records[-1].success is False
+
+
+def test_llm_request_snapshot_defaults_to_compact_payload(tmp_path: Path) -> None:
+    loop = AgentLoop(loader=EngineLoader(workspace=tmp_path, skill_assembler=StubSkillAssembler([])))
+    bundle = ProviderBundle(
+        main_runtime=SimpleNamespace(model="stub-model", provider_name="stub"),
+        main_provider=StubProvider(
+            [LLMResponse(content="done", finish_reason="stop", provider_name="stub", model="stub-model")]
+        ),
+    )
+
+    result = asyncio.run(loop.process_direct("hello", provider_bundle=bundle))
+    loaded = loop.boot()
+    events = loaded.session_manager.get_run_event_records(result.session_id, result.run_id)
+    snapshot = next(event for event in events if event.event_type == "llm_request_snapshotted")
+
+    assert "message_count" in snapshot.event_payload
+    assert "tool_names" in snapshot.event_payload
+    assert "messages" not in snapshot.event_payload
+    assert "tools" not in snapshot.event_payload