Hey peeps - question: what is the standard approach for managing self-hosted runner image customisations. Say I want to install a bunch of CLI tools in my image, do that with docker and bake the tools in the image, chain workflows together so the tools can be installed declaratively or use packages/caching? Keen to know what others have done before I go too far down the rabbit hole.

If you are on AWS, using pre-baked AMIs is probably the easiest / most efficient. If you are able to retrieve pre-built versions of your tools from a fast remote storage cache (e.g. s3 if on AWS), caching is also an option. I'm not sure what you mean by "chain workflows together so the tools can be installed declaratively or use packages/caching".

Hey @Cyril (runs-on.com) - thanks for getting back to me. I just meant a layered approach using caching, shared workflows, where "workflow A" represents installing tools or retrieving tool caches and "workflow B" represents the actual work to be done. I think you already answered my question, I've just seen a few different approaches and wanted to know what the most common pattern was.