Identifies and labels cognitive distortions in input text (e.g., catastrophizing, black‑and‑white thinking), returning types, text spans, and brief explanations.