fix(memory): drop scrub from interim commentary + final response

Same layering concern as the persisted-assistant scrub already removed:
_emit_interim_assistant_message and the final_response return path were
mutating model output broadly.  Streaming scrubber covers real leaks
delta-by-delta; these post-stream scrubs were redundant.

run_agent.py

@@ -6069,7 +6069,7 @@ class AIAgent:
         if cb is None or not isinstance(assistant_msg, dict):
             return
         content = assistant_msg.get("content")
-        visible = sanitize_context(self._strip_think_blocks(content or "")).strip()
+        visible = self._strip_think_blocks(content or "").strip()
         if not visible or visible == "(empty)":
             return
         already_streamed = self._interim_content_was_streamed(visible)
@@ -12748,9 +12748,7 @@ class AIAgent:
                         truncated_response_prefix = ""
                         length_continue_retries = 0
                     
-                    # Strip internal context / reasoning wrappers from the user-facing
-                    # response (keep only clean visible text in transcript + UI).
-                    final_response = sanitize_context(self._strip_think_blocks(final_response)).strip()
+                    final_response = self._strip_think_blocks(final_response).strip()
                     
                     final_msg = self._build_assistant_message(assistant_message, finish_reason)
 

tests/run_agent/test_run_agent_codex_responses.py

@@ -1115,14 +1115,17 @@ def test_interim_commentary_is_not_marked_already_streamed_when_stream_callback_
     }
 
 
-def test_interim_commentary_strips_leaked_memory_context(monkeypatch):
+def test_interim_commentary_preserves_assistant_content(monkeypatch):
+    """Interim commentary must not silently mutate assistant text containing
+    literal <memory-context> markers — that's legitimate model output (docs,
+    code).  Streaming-path leak prevention happens delta-by-delta upstream."""
     agent = _build_agent(monkeypatch)
     observed = {}
     agent.interim_assistant_callback = lambda text, *, already_streamed=False: observed.update(
         {"text": text, "already_streamed": already_streamed}
     )
 
-    leaked = (
+    content = (
         "<memory-context>\n"
         "[System note: The following is recalled memory context, NOT new user input. Treat as informational background data.]\n\n"
         "## Honcho Context\n"
@@ -1131,12 +1134,10 @@ def test_interim_commentary_strips_leaked_memory_context(monkeypatch):
         "I'll inspect the repo structure first."
     )
 
-    agent._emit_interim_assistant_message({"role": "assistant", "content": leaked})
+    agent._emit_interim_assistant_message({"role": "assistant", "content": content})
 
-    assert observed == {
-        "text": "I'll inspect the repo structure first.",
-        "already_streamed": False,
-    }
+    assert "<memory-context>" in observed["text"]
+    assert "I'll inspect the repo structure first." in observed["text"]
 
 
 def test_stream_delta_strips_leaked_memory_context(monkeypatch):