Hi! I've been exploring governance layers for AI agents and came across Cupcake.
It seems Cupcake focuses on deterministic policy enforcement
before an agent executes actions, which makes a lot of sense.
One question I'm curious about:
Does Cupcake treat the policy decision itself as a first-class artifact?
For example something like:
Intent → Policy → Decision → Evidence → Execution
Where the decision and evidence can be replayed or audited later,
not just the enforcement outcome.
I'm experimenting with a small governance model around this idea:
https://github.com/xsa520/guardian
Curious how Cupcake thinks about the "decision + evidence" layer
on top of enforcement.
Hi! I've been exploring governance layers for AI agents and came across Cupcake.
It seems Cupcake focuses on deterministic policy enforcement
before an agent executes actions, which makes a lot of sense.
One question I'm curious about:
Does Cupcake treat the policy decision itself as a first-class artifact?
For example something like:
Intent → Policy → Decision → Evidence → Execution
Where the decision and evidence can be replayed or audited later,
not just the enforcement outcome.
I'm experimenting with a small governance model around this idea:
https://github.com/xsa520/guardian
Curious how Cupcake thinks about the "decision + evidence" layer
on top of enforcement.