How about first trying out master and if it fails then checking out with main to fetch the gist? Maybe some try/catch blocks will handle this for us.
Or is it still that using Octokit is much bugproof solution avoiding weird cornercases?

Or is it still that using Octokit is much bugproof solution avoiding weird cornercases?

I think there are two main advantages to using Octokit like in Fiddle:

• No dependency on Git being installed on the system (simple-git requires that currently)
• Octokit can use a GitHub token to avoid rate-limiting problems

I think the rate-limiting is less of a concern when doing simple Git commands (as opposed to using the API), but I can find some examples on Google of folks getting rate-limited on git commands with GitHub.

I think removing the need for Git on the system is a strong reason to use Octokit, since it makes it easier to use fiddle-core on any machine.

Sounds good! I believe we can fetch GitHub gists without being authenticated (token) using Octokit but for better rate limits we might need a token in that case.
Should we create a new API (and new CLI arg) in fiddle-core that takes in the token if provided by the user?
(This token can be used in fiddle as well)

Should we create a new API (and new CLI arg) in fiddle-core that takes in the token if provided by the user?
(This token can be used in fiddle as well)

It's generally a good idea to avoid providing access tokens as arguments, otherwise they stay exposed in command history, and any usage on for example CI jobs would show the token. For use with the CLI it would be better to use an environment variable, something like FIDDLE_CORE_GITHUB_TOKEN.

Yup, makes sense. Working on it 👍