Normalize data agent records output path

src/agent_tool_specs/data_agent.py

@@ -135,7 +135,14 @@ def build_data_agent_tools(handlers: Mapping[str, ToolHandler]) -> list[AgentToo
                     'turn_mix': {'type': 'string', 'description': 'Single-turn and multi-turn count or ratio.'},
                     'coverage': {'type': 'string', 'description': 'Query types, intent boundaries, or error types to cover.'},
                     'exclusions': {'type': 'string', 'description': 'Negative examples or boundaries to avoid.'},
-                    'output_path': {'type': 'string', 'description': 'Where draft, records, and validation files should be written.'},
+                    'output_path': {
+                        'type': 'string',
+                        'description': (
+                            'Requested records path. The runtime normalizes canonical records to the current '
+                            'session output/records.jsonl; use output/records.jsonl and do not add dataset '
+                            'subdirectories or random names.'
+                        ),
+                    },
                     'notes': {'type': 'string'},
                 },
                 'required': [
@@ -500,7 +507,11 @@ def build_data_agent_tools(handlers: Mapping[str, ToolHandler]) -> list[AgentToo
                     },
                     'output_path': {
                         'type': 'string',
-                        'description': 'Workspace-relative path to write, usually ending in .jsonl.',
+                        'description': (
+                            'Requested workspace-relative path. For canonical records, the runtime normalizes this '
+                            'to the current session output/records.jsonl or output/records.json; do not encode '
+                            'dataset names or timestamps in the filename.'
+                        ),
                     },
                     'output_format': {
                         'type': 'string',