The way we run danger to comment on PRs requires a token. There's another way to do this where it doesn't in a 2-step workflow, see https://github.com/opensearch-project/opensearch-api-specification/blob/main/.github/workflows/pr-comment.yml for an example.
Reimplement danger that way.