Can't convert a `spec_tbl_df/tbl_df/tbl/data.frame` object to function

This question is probably about magrittr but I am not sure. If I pipe data (%>%) into the next function it goes to the first argument as far as I understand. I have this example dataset (df) and would like to extract all rows that contain the values 101 to 104 in columns col1 to col3.

# A tibble: 5 x 4
     ID  col1  col2  col3
  <dbl> <dbl> <dbl> <dbl>
     1   101   102   201
     2   201   202   203
     3   104    NA   301
     4   101    NA    NA
     5   201   301   302

I can do this

library(tidyverse)

df %>% filter(pmap_lgl(select(., starts_with("col")), ~any(c(...) %in% c(101:104))))

However, when I want to get only the boolean vector I get a warning

df %>% pmap_lgl(select(., starts_with("col")), ~any(c(...) %in% c(101:104)))
Error: Can't convert a `spec_tbl_df/tbl_df/tbl/data.frame` object to function

I figured out that I get it to work using

df %>% {pmap_lgl(select(., starts_with("col")), ~any(c(...) %in% c(101:104)))}

As far as I understand df would usually be passed to the first argument of pmap but like this it goes where the dot is. However, why is this not necessary in the first case.

The data:

df <- structure(list(ID = c(1, 2, 3, 4, 5), col1 = c(101, 201, 104, 
101, 201), col2 = c(102, 202, NA, NA, 301), col3 = c(201, 203, 
301, NA, 302)), class = c("spec_tbl_df", "tbl_df", "tbl", "data.frame"
), row.names = c(NA, -5L), spec = structure(list(cols = list(
    ID = structure(list(), class = c("collector_double", "collector"
    )), col1 = structure(list(), class = c("collector_double", 
    "collector")), col2 = structure(list(), class = c("collector_double", 
    "collector")), col3 = structure(list(), class = c("collector_double", 
    "collector"))), default = structure(list(), class = c("collector_guess", 
"collector")), skip = 1), class = "col_spec"))