Bug
In layer_audio_id.py (~line 735), Layer 1 writes AI-parsed author/title into current_author/current_title columns BEFORE validation at ~line 771. When validation rejects the author as garbage, the recovery code updates verification_layer and status but does NOT revert current_author/current_title. The garbage values persist permanently.
Impact
The original (potentially correct) author/title is irreversibly overwritten with AI pollution like publisher names or audio intro fragments. This enables the worst-case outcome: wrong renames from bad data.
Fix
Validate BEFORE the UPDATE at line 735, or add current_author/current_title reversion to the recovery UPDATEs.
Severity
Critical — directly enables wrong renames.
Found via pipeline logic audit.
Bug
In
layer_audio_id.py(~line 735), Layer 1 writes AI-parsed author/title intocurrent_author/current_titlecolumns BEFORE validation at ~line 771. When validation rejects the author as garbage, the recovery code updatesverification_layerandstatusbut does NOT revertcurrent_author/current_title. The garbage values persist permanently.Impact
The original (potentially correct) author/title is irreversibly overwritten with AI pollution like publisher names or audio intro fragments. This enables the worst-case outcome: wrong renames from bad data.
Fix
Validate BEFORE the UPDATE at line 735, or add
current_author/current_titlereversion to the recovery UPDATEs.Severity
Critical — directly enables wrong renames.
Found via pipeline logic audit.