The "GitHub PR review copilot" example will create an automation that polls every 15 mins, but it still A) runs a conversation to find new PRs, thus burning LLM tokens rather than B) running the python script and intelligently determining when a new conversation is required. Has anyone been able to get Experience B working on first try?

nit: we should drop "copilot" from this name this seems like a problem with the skill, shouldn't be hard to fix