fix(cli): pass None as system_message in manual /compress to prevent duplication (#15281)

_manual_compress was passing _cached_system_prompt to _compress_context,
but _build_system_prompt already assembles the identity block into
prompt_parts.  The redundant system_message caused the identity to
appear twice in the compressed output.  Pass None so _compress_context
rebuilds the prompt from scratch.

Supersedes #15304.

cli.py

@@ -7023,9 +7023,15 @@ class HermesCLI:
                 else:
                     print(f"🗜️  Compressing {original_count} messages (~{approx_tokens:,} tokens)...")
 
+                # Pass None as system_message so _compress_context rebuilds
+                # the system prompt from scratch via _build_system_prompt(None).
+                # Passing _cached_system_prompt caused duplication because
+                # _build_system_prompt appends system_message to prompt_parts
+                # which already contain the agent identity — resulting in the
+                # identity block appearing twice (issue #15281).
                 compressed, _ = self.agent._compress_context(
                     original_history,
-                    self.agent._cached_system_prompt or "",
+                    None,
                     approx_tokens=approx_tokens,
                     focus_topic=focus_topic or None,
                 )

tests/cli/test_manual_compress_system_prompt.py

@@ -0,0 +1,90 @@
+"""Tests for manual /compress system prompt handling.
+
+When _manual_compress invokes _compress_context, it must pass None
+as system_message to avoid duplicating the agent identity block.
+Issue #15281.
+"""
+import unittest
+from contextlib import contextmanager
+from unittest.mock import MagicMock, patch
+
+
+class TestManualCompressSystemMessage(unittest.TestCase):
+    """Verify _manual_compress avoids system prompt duplication."""
+
+    def _make_cli_with_agent(self):
+        """Create a minimal CLI mock with enough state for _manual_compress."""
+        from cli import HermesCLI
+
+        cli = MagicMock(spec=HermesCLI)
+        cli.conversation_history = [
+            {"role": "user", "content": "msg1"},
+            {"role": "assistant", "content": "resp1"},
+            {"role": "user", "content": "msg2"},
+            {"role": "assistant", "content": "resp2"},
+            {"role": "user", "content": "msg3"},
+            {"role": "assistant", "content": "resp3"},
+        ]
+        cli.agent = MagicMock()
+        cli.agent.compression_enabled = True
+        cli.agent._cached_system_prompt = "You are Hermes, a helpful AI assistant."
+        cli.agent.session_id = "test-session-id"
+        cli.session_id = "test-session-id"
+        cli.agent._compress_context.return_value = (
+            [{"role": "assistant", "content": "[compressed summary]"}],
+            50,
+        )
+
+        @contextmanager
+        def _busy_stub(msg):
+            yield
+
+        cli._busy_command = _busy_stub
+        return cli
+
+    @patch(
+        "agent.manual_compression_feedback.summarize_manual_compression",
+        return_value="Compressed 6 → 1",
+    )
+    @patch(
+        "agent.model_metadata.estimate_messages_tokens_rough",
+        return_value=5000,
+    )
+    def test_compress_context_receives_none_system_message(
+        self, _mock_tokens, _mock_summary
+    ):
+        """_compress_context must receive None, not the cached system prompt."""
+        from cli import HermesCLI
+
+        cli = self._make_cli_with_agent()
+        HermesCLI._manual_compress(cli)
+
+        cli.agent._compress_context.assert_called_once()
+        args, kwargs = cli.agent._compress_context.call_args
+        # Second positional arg is system_message — must be None
+        self.assertIsNone(
+            args[1],
+            "_compress_context should receive None as system_message, "
+            "not the cached prompt, to avoid identity duplication",
+        )
+
+    @patch(
+        "agent.manual_compression_feedback.summarize_manual_compression",
+        return_value="Compressed",
+    )
+    @patch(
+        "agent.model_metadata.estimate_messages_tokens_rough",
+        return_value=5000,
+    )
+    def test_compress_context_receives_focus_topic(
+        self, _mock_tokens, _mock_summary
+    ):
+        """Focus topic from /compress <topic> should be forwarded."""
+        from cli import HermesCLI
+
+        cli = self._make_cli_with_agent()
+        HermesCLI._manual_compress(cli, cmd_original="/compress database schema")
+
+        cli.agent._compress_context.assert_called_once()
+        _, kwargs = cli.agent._compress_context.call_args
+        self.assertEqual(kwargs.get("focus_topic"), "database schema")