How to extracting elevation values from elevatr package in R

Question

I recently started using R for spatial analysis and have some problems along the way.

I am using the elevatr package in R to make an elevation map for São Paulo state, Brazil. My idea is to use latitude, longitude and elevation values to extract climate variables from points on the map.

However I have a problem when trying to extract elevation values for points plotted on the map when using the st_crop and aggregate functions.

The code I'm using is following:

library(tidyverse)
library(sf)
library(elevatr)
library(stars)
library(rnaturalearthhires)

# São Paulo map:

brazil.est <- ne_states(country = 'Brazil', returnclass = 'sf')
sao_paulo.br <- brazil.est %>% filter(woe_name == "São Paulo")

# Random latitude and longitude points:

set.seed(123)
grid_points <- tibble(
x = runif(200, st_bbox(sao_paulo.br)[1], st_bbox(sao_paulo.br)[3]), 
y = runif(200, st_bbox(sao_paulo.br)[2], st_bbox(sao_paulo.br)[4])
)

grid_points <- st_as_sf(grid_points, coords = c("x", "y"), crs=4326)

# Selectin points that are inside São Paulo map:

grid_points2 <- grid_points_teste %>% 
     mutate(on_land=lengths(st_within(grid_points_teste,sao_paulo.br))) %>% 
     filter(on_land==1)

ggplot()+
geom_sf(data=sao_paulo.br)+
geom_sf(data=grid_points, color="#000000")+
geom_sf(data=grid_points2, color="red")

# Elevation raster from elevatr package:

elevation_sp <- elevatr::get_elev_raster(locations = sao_paulo.br, z=8, clip = 
"locations")
elevation.df <- as.data.frame(elevation_sp, xy=TRUE)
colnames(elevation.df)[3] <- "elevation"
elevation.df <- elevation.df[complete.cases(elevation.df),]

ggplot()+
 geom_raster(data=elevation.df, aes(x=x, y=y, fill=elevation))+
 geom_sf(data=sao_paulo.br, color="white", fill=NA)+
 geom_sf(data=grid_points2, color="red")

 # Extracting elevation values from points on the map using st_crop and aggregate:

 grid_pontos.cortados2 <- st_crop(grid_points2, elevation_sp, crop=T, as_points = 
 all(na.rm = T))

 sp_elevation_pontos <- aggregate(elevation_sp, grid_pontos.cortados2, mean)

And then it returns the following error even with the geometry column in the dataframe grid_pontos.cortados2:

 Error in Math.data.frame(list(on_land = c(1L, 1L, 1L, 1L, 1L, 1L, 1L,  : 
 non-numeric-alike variable(s) in data frame: geometry

Does anybody know the reason this error occurs or how to fix it? And have any tip to improve my code?