Winsorizing longitudinal dataset

I would like to winsorize my data. As data is longitudinal (see below for an example) I have to winsorize by ID. Data collection was finished when the value 100 was reached. (Below is just an example, not my real data set. I have few (1-3) outliners that reached value 100 at for example scale= 20 when the majority of people reached 100 at scale =5. Therefore, I would like to 'clean' my data.

ID scale value
CAR   1    0
CAR   2    50
CAR   3    70
CAR   4    100
RIDE   1    0
RIDE   2    0
RIDE   3    1
RIDE   4    10
RIDE   5    30
RIDE   5    40
RIDE   6    80
RIDE   7    100
BUS   1    25
BUS   2    60
BUS   3    80
BUS   4    90
BUS   5    100

I've tried

library(psych)
winsor(data$value, trim = 0.05, na.rm=TRUE, by="ID")

but it won't work. Any ideas?