This paper derives, from the two fundamental operations of Generative Geometry — hold and cross — a complete theory of intervention in any dissipative process. The theory produces ten structural results: (1) a sequential gate blockade formula with depth correction; (2) five agent functions derived from the two operations, the fractal property, and the cyclic nature of the intervention itself; (3) an opacity formula derived from position 13 (Differentiation); (4) intelligence defined as observation depth, derived from the opacity structure; (5) four intervention strategies named by intervener intent — Dissolution, Disruption, Rejection, Occupation; (6) the mutual arising of depth at the singularity threshold between Conservation and Potentiality; (7) a regime-labelled intervention formula whose four parameters each correspond to one of the four regimes, each derived from first principles; (8) a fractal extension resolving within-phase agent combinations through sub-level coverage; (9) two-component synergy — σ_cascade from sequential dependency and σ_generative from the generative property (Φ > 0) of the intervention encounter, explaining why some drug combinations produce dramatically more than structural coverage predicts; and (10) eight intervention actions (two per sub-phase) that determine the agent's structural effect beyond potency. With per-agent calibration from published single-agent data, the fractal formula with two-component synergy produces 0.2% mean absolute error across eight breast cancer combinations, 0.6% across nine melanoma combinations, and 0.0% across three NSCLC combinations. Across 25 predictions with calibrated parameters, mean absolute error is 0.6%. v2 addendum (March 2026) connects the blockade formula to the Four-State Observer Protocol (10.5281/zenodo.19130570), identifying the intervention formula as T1 (load reduction) within a four-transition framework for durable remission.