Severe mental illness (SMI) is used here as a functional and service-relevant category: mental disorders associated with substantial, persistent, or recurrent restrictions in major life activities, role participation, treatment stability, or safety. Psychotherapy in SMI requires more than symptom reduction. Clinicians must repeatedly decide whether to deepen emotional work, downshift to regulation, address avoidance, protect emerging progress, or intervene against shame-driven self-attack. These decisions are clinically central but often remain poorly documented and difficult to supervise, compare, or validate. This Hypothesis & Theory manuscript proposes an SMI-specific clinical translation of the Perceive-Assess-Dose-Safeguard (PAD-S) decision matrix and the related Conflict-Square Algorithm (CSA) as a sparse, human-final decision language for such moments. PAD-S/CSA organizes session cues into four process nodes: Progression (PRO), Anxiety/Affect tolerance (ANX), Defense/Avoidance (DEF), and Superego/Shame attack (SUP). Each node is linked to tolerance thresholds, calibrated therapist moves, safeguards, and expected functional impact. The framework connects three levels without collapsing them: the International Classification of Diseases, 11th Revision (ICD-11), together with the Clinical Descriptions and Diagnostic Requirements (CDDR), anchors diagnosis and severity context; the International Classification of Functioning, Disability and Health (ICF) and the Mini-ICF-APP rating for activity and participation restrictions in mental disorders anchor activity and participation; PAD-S/CSA documents session-level process decisions. The manuscript clarifies novelty relative to related PAD-S/CSA publications, defines SMI operationally, provides plain-language node definitions, gives an SMI-specific adaptation table, illustrates a worked micro-sequence and two brief vignettes, and proposes a staged validation agenda. Digital extensions are framed as optional annotation, summarization, and research support; clinical responsibility remains human-final.