fix: 自动注册读脚本docstring推算数据流，不再是"未知"

scripts/morning_health_check.py

@@ -647,26 +647,55 @@ def run_check(item):
             if unregistered:
                 ok = False
                 detail = f"{len(gaps)}条待验证 + {len(unregistered)}个新组件未注册"
-                # 自动修复：把未注册的脚本加入注册表（占位）
+                # 自动修复：读脚本docstring，推算数据流
                 try:
                     for s in unregistered:
+                        script_path = HERMES_CRON_DIR.parent / "scripts" / f"{s}.py"
+                        desc = "未知"
+                        source_info = f"{s}.py"
+                        consumer_info = "未知"
+                        if script_path.exists():
+                            content = script_path.read_text()
+                            # 提取docstring
+                            import re as rr2
+                            doc_match = rr2.search(r'"""(.*?)"""', content, rr2.DOTALL)
+                            if doc_match:
+                                doc_text = doc_match.group(1).strip()
+                                desc = doc_text.split('\\n')[0][:80]
+                                # 尝试从docstring中提取管道信息
+                                pipe_match = rr2.search(r'管道[：:].*?(?=\\n|$)', doc_text)
+                                if pipe_match:
+                                    consumer_info = pipe_match.group(0).replace('管道','').strip('：: ')
+                            # 检测写入模式
+                            if 'signal_news' in content:
+                                consumer_info = 'signal_news表'
+                            if 'macro_risk_state' in content:
+                                consumer_info = 'macro_risk_state.json'
+                            if 'watchlist' in content.lower():
+                                consumer_info = 'watchlist.json / decisions.json'
+                            if 'INSERT INTO' in content:
+                                for tbl in ['todos', 'price_events', 'macro_context_log', 'accuracy_stats']:
+                                    if tbl in content:
+                                        consumer_info = f'{tbl}表'
+                            if '.write_text' in content or 'json.dump' in content:
+                                for path in ['macro_risk_state', 'macro_context', 'market', 'portfolio', 'decisions']:
+                                    if path in content:
+                                        consumer_info = f'{path}.json'
                         reg["pipelines"].append({
                             "id": f"auto-{s}",
-                            "name": f"{s}(自动发现)",
-                            "source": f"{s}.py (自动发现)",
-                            "consumer": "未知(需确认)",
-                            "end_user": "未知(需确认)",
+                            "name": desc[:60],
+                            "source": source_info,
+                            "consumer": consumer_info,
+                            "end_user": "待确认",
                             "verified": False,
-                            "verified_at": None,
-                            "gap": "自动发现的组件，数据流待确认",
-                            "fix": "手动编辑pipeline_registry.json完善此项",
-                            "deadline": "发现后3天内"
+                            "gap": f"自动发现({desc[:60]})",
+                            "fix": "手动编辑pipeline_registry.json完善此项"
                         })
                     open(str(DATA / "pipeline_registry.json"), 'w').write(
                         j2.dumps(reg, ensure_ascii=False, indent=2))
-                    detail += f" → 已自动添加{len(unregistered)}条占位记录"
-                except Exception:
-                    pass
+                    detail += f" → 已自动注册{len(unregistered)}个(含推断)"
+                except Exception as e:
+                    detail += f" (自动注册异常:{str(e)[:30]})"
             elif gaps:
                 ok = False
                 detail = f"{len(gaps)}条管道未验证: {', '.join(gaps[:5])}"