What did your AI-built app miss before deploy?

[ohad6k]

AI-built apps often look done locally and then fail on the boring launch layer.

I am collecting concrete examples VibeRaven should learn to catch from local repo evidence.

Useful replies look like this:

1. Stack: Next.js + Supabase + Vercel, Clerk + Vercel, Stripe, Polar, Railway, etc.
2. What passed locally?
3. What failed in preview or production?
4. What repo files, env names, scripts, migrations, routes, or tests would have hinted at the problem?
5. Should this be a repo-code fix, a provider-dashboard action, or a user confirmation?

Examples I am especially interested in:

• Supabase returns empty data because RLS policy evidence is missing or wrong.
• Auth works locally but preview/prod callback URLs drift.
• Webhook route exists but does not verify signatures.
• Env vars exist only in .env.local and not in deploy docs/config.
• Monitoring, Sentry, PostHog, or deploy checks are claimed but not evidenced.

Please do not post secrets, tokens, private keys, signing secrets, real customer data, or live provider credentials.

If the report is specific enough to track as work, open an issue too:

• False positive: https://github.com/ohad6k/VibeRaven/issues/new?template=false-positive.yml
• Missed production gap: https://github.com/ohad6k/VibeRaven/issues/new?template=missed-gap.yml
• Stack support request: https://github.com/ohad6k/VibeRaven/issues/new?template=stack-request.yml

Open the local launch console from an app repo:

npx -y viberaven

For the deeper non-interactive agent/pro gate:

npx -y viberaven --agent-mode