diff --git a/navi/core/subagent_runner.py b/navi/core/subagent_runner.py
index 1c6c202..23dc34a 100644
--- a/navi/core/subagent_runner.py
+++ b/navi/core/subagent_runner.py
@@ -207,7 +207,8 @@
 
             for iteration in range(max_iterations):
                 if stop_event and stop_event.is_set():
-                    return accumulated_text, False
+                    report = self._build_progress_report(context, "user_stop", iteration)
+                    return report + "\n\n" + (accumulated_text or ""), False
 
                 elapsed = time.monotonic() - _start_time
                 if elapsed >= timeout_seconds:
@@ -220,7 +221,8 @@
                                 token_count=_turn_tokens, tool_call_count=_sub_tool_count
                             )
                         )
-                    return accumulated_text or "[Sub-agent timed out]", False
+                    report = self._build_progress_report(context, "timeout", iteration)
+                    return report + "\n\n" + (accumulated_text or "[Sub-agent timed out]"), False
 
                 log.debug("agent.subagent.iteration", iteration=iteration)
 
@@ -293,7 +295,8 @@
                                 tool_call_count=_sub_tool_count,
                             )
                         )
-                    return f"[{thinking_stalled_reason}]", False
+                    report = self._build_progress_report(context, "thinking_stall", iteration + 1)
+                    return report + "\n\n" + f"[{thinking_stalled_reason}]", False
 
                 if not turn_tool_calls:
                     log.info(
@@ -308,7 +311,13 @@
                                 tool_call_count=_sub_tool_count,
                             )
                         )
-                    return accumulated_text, True
+                    report = self._build_progress_report(context, "completed", iteration + 1)
+                    result = accumulated_text or ""
+                    if result:
+                        result = report + "\n\n" + result
+                    else:
+                        result = report
+                    return result, True
 
                 if accumulated_thinking and sink is not None:
                     log.debug(
@@ -341,7 +350,8 @@
                     # skip remaining tools in this batch.
                     if stop_event and stop_event.is_set():
                         log.info("agent.subagent.tool_batch_stopped", tool=tc.name)
-                        return "", False
+                        report = self._build_progress_report(context, "user_stop", iteration + 1)
+                        return report + "\n\n", False
 
                     _sub_tool_count += 1
                     if sink is not None:
@@ -424,11 +434,9 @@
                         token_count=_turn_tokens, tool_call_count=_sub_tool_count
                     )
                 )
-            return (
-                accumulated_text
-                or "[Sub-agent reached iteration limit without a final answer]",
-                False,
-            )
+            report = self._build_progress_report(context, "max_iterations", max_iterations)
+            result = accumulated_text or "[Sub-agent reached iteration limit without a final answer]"
+            return report + "\n\n" + result, False
         finally:
             _sid_var.set(_prev_sid)
             _model_var.set(_prev_model)
@@ -436,5 +444,64 @@
             _role_var.set(_prev_role)
             _uinfo_var.set(_prev_uinfo)
 
+    @staticmethod
+    def _build_progress_report(
+        context: list[Message],
+        reason: str,
+        iterations_used: int,
+    ) -> str:
+        """Serialize the sub-agent's execution history into a progress report.
+
+        The report lists every tool call the sub-agent made, the arguments,
+        and a short result snippet so the parent agent knows what was
+        accomplished before the sub-agent stopped.
+        """
+        lines = [
+            f"[Sub-agent stopped: {reason}]",
+            f"Iterations used: {iterations_used}",
+            "",
+        ]
+
+        turn_number = 0
+        msg_idx = 0
+        while msg_idx < len(context):
+            msg = context[msg_idx]
+            if msg.role == "assistant" and msg.tool_calls:
+                turn_number += 1
+                turn_lines = [f"Turn {turn_number}:", ""]
+                if msg.content:
+                    text = msg.content.strip().replace("\n", " ")
+                    turn_lines.append(f"  Assistant text: {text[:200]}{'...' if len(text) > 200 else ''}")
+                for tc in msg.tool_calls:
+                    # Find matching tool result
+                    result_msg = None
+                    for m in context[msg_idx + 1 :]:
+                        if m.role == "tool" and m.tool_call_id == tc.id:
+                            result_msg = m
+                            break
+                    if result_msg:
+                        status = "error" if result_msg.content.startswith("Error:") else "success"
+                        snippet = result_msg.content.strip().replace("\n", " ")[:120]
+                        turn_lines.append(f"  - {tc.name} → {status}: {snippet}{'...' if len(result_msg.content) > 120 else ''}")
+                    else:
+                        args = str(tc.arguments).replace("\n", " ")[:80]
+                        turn_lines.append(f"  - {tc.name}({args}) → (no result yet)")
+                lines.extend(turn_lines)
+                lines.append("")
+                msg_idx += 1 + len(msg.tool_calls)
+            elif msg.role == "assistant" and msg.content:
+                turn_number += 1
+                text = msg.content.strip().replace("\n", " ")
+                lines.append(f"Turn {turn_number}: Assistant response: {text[:200]}{'...' if len(text) > 200 else ''}")
+                lines.append("")
+                msg_idx += 1
+            else:
+                msg_idx += 1
+
+        if turn_number == 0:
+            lines.append("No tool calls recorded.")
+
+        return "\n".join(lines)
+
     def _get_backend(self, backend_key: str) -> LLMBackend:
         return self._backends.get(backend_key)
diff --git a/tests/unit/core/test_agent.py b/tests/unit/core/test_agent.py
index e48fa6c..022e44d 100644
--- a/tests/unit/core/test_agent.py
+++ b/tests/unit/core/test_agent.py
@@ -224,7 +224,8 @@
         agent._backends.register("ollama", backend)
 
         result, ok = await agent.run_ephemeral("task", profile_id="test")
-        assert result == "subagent result"
+        assert "subagent result" in result
+        assert "[Sub-agent stopped: completed]" in result
         assert ok is True
 
     @pytest.mark.asyncio