Every now and then it is worth reviewing your code, to see if the architecture, guardrails, security, speed, etc are as good as they can be. AI dev is amazing, but it does occasionally go off the rails.
I decided to do a major review of my entire codebase while testing out the “High” power mode in the new agent 3-va (variable autonomy) - kill two birds with one stone as they say 
Run your own app code audit
Below is the single prompt I used - you are welcome to borrow it for your apps.
Importantly, the second paragraph needs to be modified so it is specific to your app. This gives the agent a short intro that explains what your app is supposed to do - so agent isn’t just blindly looking at code without better context. If you have any specific and important scripts as I did, mention these so it can put a little more focus on them:
I have put you into “High” autonomy mode to do a comprehensive analysis and review of my entire codebase, looking for issues and concerns, including security flaws. You will not make changes, this is just a review. Before you start, read the readme and replit.md in-depth to understand more about this app. In addition, you should do your analysis with the following app outline in mind:
The app is my AI Dev Kit (ADK), essentially a scaffold app to let other Replit users remix and build an app with a core foundation. More information can be found at my website gipity.com. The ADK is heavily opinionated in some of the architecture, mainly because my target Replit users are non-tech startup founders building MVPs, which I help them with in my Support Hub. As part of this, the ADK uses Supabase for auth, DB and file storage (the gipity-supabaseprepare.sql script is used to initialise Supabase). It is also designed to build web, PWA and native iOS/Android apps from a single codebase. gipity-appflow-prepare.sh and gipity-image-resizer.py are key scripts you should understand, as they are used as part of the native app-building workflow, together with Ionic Appflow.
Total cost $2.38. Once we had agreed on the tasklist, it ran for just under 10 minutes.
The report it gave was amazing and incredibly detailed. Broken down into useful and very readable sections. The kind of thing an analyst would take 2 days to carry out and write up - at a price much higher than $2.38 
I won’t show my report here (unless someone really wants to see it). But I am pleased to say my app passed it’s review with flying colours 
And, agent 3-va(high) worked really smoothly.
Let me know if you try it, and the kind of results you get - fingers crossed Replit is starting to produce some good code!