sigh finally refactored a load of legacy 🗑️ to the point of it being usable w/ pants got everything tied in. CI scripts only build and deploy what's changed based on a tag... First deploy goes well... Final step fails updating the tag because gitlab doesn't let CI mutate the repository. 🤦‍♂️ So the next build will just rebuild and redeploy... everything.

In my experience you can just create an automation gitlab user, create a token for it, and inject that in to the CI process via environment and use that to user to manipulate the repository however you want

This seems to be what the previous incumbents have done.

gitlab automatically injects some config and env vars into the CI job, I forget how exactly, but you can always manually inject your own