The power flexibility of the proliferating distributed energy resources located in distribution networks could be aggregated and provided to the transmission system. Considering the lack of a framework to help transmission system operator (TSO) to benefit from these potential power flexibility resources, first, the article establishes a framework for modeling aggregated flexibilities of distribution networks (AFENs) as seen from TSO's perspective. Then, it develops a two-stage linear stochastic optimization model to optimally book the TSO's required size of active power flexibility from AFENs and dispatchable power plants (DPPs). The model leverages a cost-benefit method and a dc load-flow model to minimize the TSO's total cost, namely the sum of 1-expected cost of active power flexibility allocation from AFENs and DPPs and 2-expected cost of energy not supplied. To achieve a risk-aware economic balance between these two incurred costs of TSO, the model relies on the value of lost load index as a metric. The method considers credible contingencies along with forecast errors of renewable generation and loads as scenarios. Finally, the method is applied to the transmission grid of Switzerland, operated by Swissgrid, to illustrate its effectiveness.