No idea. We'll have to test. Also, I think this won't have the GITHUB_TOKEN value available in the script.

Can we do the approach of 882e4a1?

Testing that, it seems to work:

• ee7/exercism-nim@ffad676
• https://github.com/ee7/exercism-nim/actions/runs/3854842218/jobs/6569260535

Yes, that works, but I'd be most interested in whether it use the github_token (I expect it doesn't). Could you test that?

It was intentionally not meant to use the track's fetch-configlet script for stability and simplicity (no need for all the OS handling if the environment is known). Once it can be ensured that the track's script exists, isn't broken, and is always up-to-date, this would be a low-risk change, but that's not the case yet afaik.

@SaschaMann It is guaranteed to be present: https://github.com/exercism/org-wide-files/blob/main/tracks-files/bin/fetch-configlet

Merging of org-wide-files syncs isn't automatic, you'd still run into situations where the script might be out of date. For a CI script, that seems like an issue.

For now, let's close in favor of #107.

One advantage of using gh here is that it's guaranteed to fail if we don't pass the token in correctly. With the curl version, we might only notice that we're making unauthenticated requests when we hit rate limits.