Guardrails and governance

[Frits1980]

I can't find anything on how to set guardrails or governance rules on any level in the app. Is this something that is already on some roadmap? In my former harness I hade a lot of claude code hooks set up to keep AI within my boundaries of trust, but also to make the outcome as deterministic as possible (not possible, but it did help a little). I'm kinda missing that in paperclip right now. I love paperclip, in fact I think this is as huge as Lovable was in the first few months. But in my eyes it needs more fine tuning possibilities on all levels (not only the AGENT.md level). And it needs rules harder then instructions, it needs things like hooks that are impossible to work around. How do you all think of this?

[Hardrada]

I use projects to implement workflows, hand-offs to other agents and guardrails such as "you can never create an email account", "you can never sign up for .XYZ".. So you can have some global rails in agents.md or you can have some more finely tuned, per project rails that cannot supersede the globals. That's how I've doing it.