fix: recognize plain speaker lines in discussions

automation/workflow.py

@@ -265,6 +265,12 @@ def _extract_participant(line: str) -> tuple[str | None, str]:
             candidate = parts[0].strip()
             if candidate:
                 return candidate, parts[1].strip()
+    colon_pos = stripped.find(":")
+    if colon_pos > 0:
+        candidate = stripped[:colon_pos].strip()
+        remainder = stripped[colon_pos + 1 :].strip()
+        if candidate:
+            return candidate, remainder
     return None, line