This article focuses on addressing the problem of enhancing the effectiveness of natural science education in schools, particularly in physics, by applying the principle of problem-based learning and designing lessons centered on experimental activity. The foundational unit of this model is a student’s action that is motivated, deliberate, and aimed at attempting to solve a task, exploring solutions, and engaging in reflection and generalization of goal-directed actions. However, assessing the effectiveness of such actions requires high-quality research instruments capable of collecting reliable data. To address this challenge, we developed an assessment instrument and empirically tested some psychometric requirements. We hypothesized that the key requirements for such an instrument are: (1) alignment with the structure of activity; (2) multifunctionality and robustness against changes in the evaluator. These requirements were formulated by the authors based on V.V. Davydov’s theory of learning activity and the theory of psycho-pedagogical measurement. Data collection was conducted through non-participant observation, with student activity evaluated by independent experts—assessors—using audio and video recordings of experimental lessons in three schools in Yuzhno-Sakhalinsk. The validation of the instrument’s requirements was carried out through the analysis of expert judgments and the application of consistency metrics, such as the proportion of matching results and Kendall’s nonparametric concordance coefficient. The analysis of the collected data demonstrated that: (a) the developed instrument effectively captures actions related to task-solving in accordance with the theoretical model; (b) the assessments by independent observers yielded similar results, as reflected in acceptable consistency metrics for their evaluations.