Buildings represent a promising flexibility source to support the integration of renewable energy sources. Indeed, they may shift their heating energy consumption without impacting users' comfort. Literature proposes some metrics to quantify the future flexibility of buildings' heating systems but either omits or oversimplifies the impact of future uncertain outdoor conditions and model accuracy. This paper introduces an uncertainty-aware flexibility envelope, representing the minimum and maximum energy consumption of the building's heating system over the next 24 hours. The proposed chance-constrained linear optimization accounts for uncertainties about future weather conditions and modelling errors. Uncertainty-aware envelopes are compared to their uncertainty-ignorant counterparts, in which forecasts and models are assumed to be perfect. Results indicate a significant reduction of the flexibility potential when accounting for uncertainties, with the model error being the most impactful factor. Additionally, the case study presented in this paper reveals the existence of a maximum flexibility provision duration that may restrict a building's uncertainty-aware flexibility potential.