kashvats/p_assistant ? reverse-engineered prompt
Reverse engineered prompt
Build me a local first personal assistant for Windows, Ubuntu, Linux, and macOS that helps with reminders, routines, project supervision, monitoring, and basic defensive security, while keeping things lightweight until deeper reasoning is actually needed.
I want it to work through a CLI, a local API, and a simple dashboard, with the assistant storing its own state locally and being careful about secrets, approvals, and sensitive files. Please include the experience learning layer so it can remember useful lessons from tool runs, failures, and successful fixes, but keep those memories bounded and safe so they can’t override policy or trust raw external output.
Make it security hardened by default, especially around shell commands, database access, localhost API requests, browser or network access, and self improvement promotion. Look up current docs online if you need to, and then wire up the app so it runs cleanly from the repo with clear setup instructions and sensible defaults.
Are you gonna build this?
make sure you review the code using arcumet