docs(governance): clarify OpenAI Agents tool-name + duck-typing rationale

AFTER_TOOL reads the tool name directly from the SDK-provided tool at
both on_tool_start and on_tool_end, so no start→end correlation is
needed (addresses the #899-style tool-identity review note). Also record
that the run-item/response text extraction is duck-typed intentionally,
since the SDK shapes are not stable public models.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: aditik0303

packages/uipath-openai-agents/src/uipath_openai_agents/governance/adapter.py

@@ -268,7 +268,12 @@ async def on_tool_start(self, context: Any, agent: Any, tool: Any) -> None:
     async def on_tool_end(
         self, context: Any, agent: Any, tool: Any, result: Any
     ) -> None:
-        """Evaluate AFTER_TOOL rules immediately after a tool is invoked."""
+        """Evaluate AFTER_TOOL rules immediately after a tool is invoked.
+
+        The SDK passes ``tool`` to both ``on_tool_start`` and ``on_tool_end``,
+        so the name is read directly here — no start→end correlation is needed
+        (unlike callback frameworks whose end hook omits the tool).
+        """
         try:
             tool_name = getattr(tool, "name", None) or "unknown"
             tool_result = "" if result is None else _stringify(result)
@@ -301,6 +306,10 @@ async def on_handoff(self, context: Any, agent: Any, source: Any) -> None:
 
 # --------------------------------------------------------------------------
 # Delegation + text extraction (module-level, sync, duck-typed)
+#
+# Extraction is duck-typed on purpose: the OpenAI Agents SDK's run-item /
+# response shapes are not stable public models, so we read attributes
+# defensively rather than isinstance-checking SDK types that may move.
 # --------------------------------------------------------------------------