Abstract: The development of digital discipline inspection and supervision has driven the deep integration of algorithmic technologies into the exercise of supervisory power, transforming them from ordinary auxiliary tools for case handling into structural forces that shape factual screening, risk prioritization, lead triage, and procedural routing. Algorithms are not independent holders of public authority and cannot replace supervisory organs in making final disposition decisions. Yet through data aggregation, model-based computation, and workflow control, they reshape the conditions under which supervisory power operates, thereby generating algorithmic power with de facto controlling effects. The central difficulty in bringing smart supervision under the rule of law lies not merely in technical errors or data bias, but in the dysfunction of the supervisory accountability system caused by insufficient algorithmic explainability. Because supervisory power is confidential, disciplinary, and highly sensitive, the governance of algorithms in smart supervision cannot simply adopt a transparency model based on full public disclosure. Instead, within the rule-of-law framework for supervision, it is necessary to establish hierarchical and differentiated duties of explanation, a full lifecycle allocation of responsibility, closed or internal algorithmic auditing, and specialized dispute resolution and remedial mechanisms. Only by incorporating the points of algorithmic intervention, their grounds, operational pathways, and degree of influence into an institutional structure that is explainable, reviewable, and accountable can the efficacy of digital anti-corruption be unleashed while preserving the minimum requirements of statutory procedure and rights protection.