I've pushed an update that hopefully addresses these concerns. (I rebased to master - hopefully that doesn't make the diffs too messy.)

After making a proper unit test for independent sets, I found I had to convert AbstractGraph to SimpleGraph in order to call complement. Performance shouldn't be an issue since this data copy is negligible in comparison to the cost of the maximal_cliques call. But I don't know if there is some other reason this would be problematic.

@traitfn function maximum_independent_set(
    g::AG::(!IsDirected)
) where {T,AG<:AbstractGraph{T}}
    # Convert to SimpleGraph first because `complement` doesn't accept AbstractGraph.
    return maximum_clique(complement(SimpleGraph(g)))

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.22%. Comparing base (5eaa071) to head (29c01d3).
⚠️ Report is 1 commits behind head on master.

@@           Coverage Diff           @@
##           master     #155   +/-   ##
=======================================
  Coverage   97.22%   97.22%           
=======================================
  Files         122      122           
  Lines        7240     7250   +10     
=======================================
+ Hits         7039     7049   +10     
  Misses        201      201           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

I would add a warning to the docstring pointing to the conversion, but otherwise I'd say it's okay

This looks good, and thankfully it has not bitrotten in the last two years. Thank you for the contribution!