modelsummary: Robust standard errors and significance starts

When using modelsummary and clustered SEs via the vcov argument with a glm model, I get the following error:

The `lmtest::coeftest` function does not seem to produce complete results when
applied to a model of class glm. Only the standard errors have been adjusted. p
values and confidence intervals may not be correct.

And as the error says, in the resulting table only the SEs have been corrected. The significance stars (*, **, etc), which I presume are calculated from the p-values, are not.

MWE:

library(modelsummary)

df = data.frame(
  outcome = rbinom(1000, 1, 0.2),
  cov1 = rnorm(1000),
  cov2 = rnorm(1000),
  cat = rep(c("A", "B", "C", "D", "E"), each = 200))

modellist = list(
  glm(outcome ~ cov1 + factor(cat), data = df, family = "binomial"),
  glm(outcome ~ cov1 + cov2 + factor(cat), data = df, family = "binomial"))

modelsummary(models = modellist, estimate = "{estimate}{stars}", vcov = ~cat)

modelsummary(models = modellist, estimate = "{estimate}{stars}")

Is there any workaround to present the correct stars using the adjusted SEs in this case?

EDIT (solution):

It turns out that, at least when uploading to R 4.x.x, the lmtest package needs to be installed independently of modelsummary. Once that it's done, the issue disappears.