diff --git a/agent.py b/agent.py
index 0324a0b..5f78e8c 100644
--- a/agent.py
+++ b/agent.py
@@ -358,14 +358,17 @@ def _parse_inbox(inbox_text: str) -> list[dict[str, str]]:
     normalized = inbox_text.replace("\r\n", "\n").strip()
     blocks = [block.strip() for block in re.split(r"\n\s*---+\s*\n", normalized) if block.strip()]
     if len(blocks) == 1:
-        # Also split common pasted inbox format with repeated From: headers.
+        # Single unstructured paragraphs should ask for a clearer paste instead of
+        # inventing a sender/subject. Repeated From: headers are still accepted.
+        if not re.search(r"(?im)^\s*(from|sender)\s*:", normalized):
+            return []
         blocks = [b.strip() for b in re.split(r"(?=\n?From\s*:)", normalized) if b.strip()]
     out: list[dict[str, str]] = []
     for idx, block in enumerate(blocks, start=1):
-        sender = _field(block, "from") or _field(block, "sender") or "unknown@example.com"
-        subject = _field(block, "subject") or f"People inbox message {idx}"
+        sender = _field(block, "from") or _field(block, "sender")
+        subject = _field(block, "subject")
         body = _field(block, "body") or _strip_headers(block)
-        if len(body.strip()) < 8 and subject.startswith("People inbox message"):
+        if not sender or not subject or len(body.strip()) < 8:
             continue
         out.append(
             {