# Complete constructed signal decisions

Six supplied observations are fictional. No account setting or recipient was inspected, and no message was sent.

S1: ordinary member message supplied as Sent, without a read photo; receipt preference unknown. Decision sent_without_read_indicator_reading_unknown. Missing photo does not prove unread or ignored.
S2: the owner's receipt preference is supplied on and active-status audience no_one. Decision two_distinct_supplied_preferences. They concern different disclosures and can be recorded together. This is not a real saved-setting test or a claim that hiding status failed.
S3: member InMail with no read indicator for its sender. Decision inmail_sender_exception_no_read_inference. Apply the documented exception; do not invent rejection or an unread claim.
S4: message request not accepted, no read indicator. Decision pending_request_no_read_inference. The receipt boundary is relevant, but actual reading remains unknown.
S5: unseen and unread counts are supplied as 3 and 2. After the declared Messaging-icon review, they are 0 and 2. Decision unseen_cleared_unread_still_separate. A zero red unseen badge is not a zero unread count or completed response. These chosen values are not a captured inbox. The German indicator page has a wording discrepancy in its red-badge sentence; its unseen heading, final note and English page support the distinction.
S6: ordinary message with a read profile-photo indicator supplied. Decision supplied_read_indicator_understanding_unknown. Preserve that declared display; understanding, an agreement to reply and customer interest remain unknown.

Run python3 check.py after extracting with Python 3 already available. It checks supplied file identities and recomputes outcomes. Duplicate cases, another product, negative counters, disappearing unread count substituted into S5, unread/understanding claims and an unexpected InMail receipt are rejected. The blank record keeps actual settings, message type and separate observations private. Fixture results are neither native proof nor reader/customer evidence.