diff --git a/agent/evolution/executor.py b/agent/evolution/executor.py
index 63380343..5c418f3c 100644
--- a/agent/evolution/executor.py
+++ b/agent/evolution/executor.py
@@ -424,6 +424,11 @@ def run_evolution_for_session(
             enable_skills=True,
             runtime_info=getattr(agent, "runtime_info", None),
         )
+        # Mark this as a restricted review agent so runtime MCP reconciliation
+        # (ToolManager.sync_mcp_into_agent) will NOT silently re-inject MCP tools
+        # that _select_tools()/_guard_tools() intentionally withheld. Without this
+        # flag the review boundary would be re-opened on the first LLM turn.
+        review_agent._evolution_restricted = True
         # Reuse the live model so it follows the user's configured model.
         review_agent.model = agent.model
         # Inject the evolution task brief AFTER the full system prompt: the agent
diff --git a/agent/tools/tool_manager.py b/agent/tools/tool_manager.py
index 73b76943..823e8b8e 100644
--- a/agent/tools/tool_manager.py
+++ b/agent/tools/tool_manager.py
@@ -523,6 +523,16 @@ class ToolManager:
         if agent is None or not hasattr(agent, "tools"):
             return ([], [])
 
+        # Never re-inject MCP tools into a restricted Self-Evolution review agent.
+        # The review agent is created with a deliberately reduced, workspace-guarded
+        # toolset; silently re-adding configured MCP tools here would bypass that
+        # policy boundary (see agent/evolution/executor.py). The flag may live on
+        # the agent itself (Agent) or on the wrapping stream executor's .agent.
+        if getattr(agent, "_evolution_restricted", False) or getattr(
+            getattr(agent, "agent", None), "_evolution_restricted", False
+        ):
+            return ([], [])
+
         from agent.tools.mcp.mcp_tool import McpTool
         current = self._mcp_tool_instances
         registry_names = set(current.keys())