This is a really timely project. Given the pace at which teams are trying for the adoption of these AI tools, integrating something like openACA directly into Claude Code makes a ton of sense.
Couple of quick questions..
Are the security rules evaluated locally on the machine, or does it require an external API check before a tool call is allowed?
How do you handle the balance between blocking unsafe commands and not getting in the developer's way with too many false alarms?