At what point do you trust an AI coding agent with production DB migrations?

Genuinely curious how others here draw this line. I've been letting an AI

coding agent (Claude Code) handle a growing share of day-to-day

engineering on a side project — not autocomplete, actual multi-step

changes across the codebase, including schema migrations on a live

production database.

What made that feel safe enough to actually do:

- It states what it's about to touch before touching it, so a bad plan

gets caught before execution, not after.

- Rules written once in a CLAUDE.md file (things like "always ask before

a destructive migration") actually stick across sessions instead of

needing to be re-explained every time.

- I still read every migration before it runs — the trust is in "won't

surprise me," not "never needs review."

Separately, I've also been leaning hard into OTA-first shipping (Expo +

`eas update`) specifically so JS/UI fixes don't need to wait on app-store

review at all — that's been more load-bearing than I expected, for

reasons unrelated to convenience.

Curious what stacks/workflows others here are running for solo or

small-team AI-assisted development — especially anyone else who's handed

an agent real production access and how you scoped the guardrails. (For

context, the project this is all happening on is a pet-health app I'm

building — happy to share more if anyone's curious, just didn't want this

post to be about that.)