Jules: Please share your workflow when working with the Jules AI agent

[cbenhagen]

Hi

I would love to hear how you are using the Jules AI coding agent in this iOS / macOS repository. Is there a way it can spin up a macOS VM? Or how do you handle testing? Do you completely rely on GitHub actions?

/cc @ncooke3 as you have by far the most commits co-authored with Jules.

[ncooke3]

Hi @cbenhagen,

macOS VMs are currently not supported (https://jules.google/docs/environment/), so few testing workflows can be dispatched to Jules. We instead rely on GitHub Actions and/or local testing to verify the work.

So far, I've mainly used Jules for completing tasks that involve a series of repetitive changes. I find it does well on these tasks.

An agents.md file was recently checked in that Jules can reference (and contribute to).

[cbenhagen]

@ncooke3 the agents.md file was the reason I asked. I somehow hoped you had access to secret macOS VMs as you were using Jules in an Apple repository. I gave it a try to implement macOS support for TOTP which I though might be an ideal task but without a macOS VM to run the tests it got annoying very quickly so I ended up doing #15112 with Windsurf.