The CyberKnife, a renowned non-invasive stereotactic radiotherapy technique, has gained widespread adoption for its effective treatment of various cancers. The success of CyberKnife treatment relies significantly on the accurate placement of fiducial markers. This study introduces a Fiducial marker placement planning algorithm, specifically for superficial tumors that are located 20-50 mm beneath the epidermis. Three patients with treatment sites of thymus, breast, and submandibular gland were retrospectively selected for this study. The algorithm proposes numerous potential implantation sites by constructing and optimizing a B-spline surface. The site selection is predominantly based on the distance between markers, the angles among them, and the necessity for all fiducial markers to be visible in 45° oblique views without overlap. By leveraging the kd-tree, we enhanced spatial querying and addressed the challenge using graph theory, specifically through the Bron-Kerbosch algorithm. The time complexity of the algorithm proposed in this study is \(O(m\log m + m^2 + 3^{\frac{n}{3}})\), significantly faster than the brute-force \(O(n^3)\) approach. Experimental outcomes show that our algorithm can efficiently plan Fiducial marker placements, simplifying the planning process and providing valuable technical support for CyberKnife treatments.