SafeAgentDB

SafeAgentDB is an Agent Skill that sets up database safety infrastructure for teams where people and AI agents work on many branches at once. Each branch and PR gets a live preview URL backed by its own isolated database, so agents can run migrations, load realistic data and test real app behaviour without risking production or shared development data. What you can do with it - Inspect your project and propose a setup, written up as a database branching doc you review before anything changes - Run a local Docker database, a persistent develop/staging database, and isolated preview databases for feature and agent branches - Wire Vercel preview env vars to each preview database and redeploy previews after env changes - Keep feature migrations on their previews until merge, check migrations before merge and deploy production migrations from main - Optionally copy develop data, auth users and storage objects into previews - Clean up preview databases when PRs close, plus a scheduled cleanup of orphaned previews - Dry-run preview provisioning before the first real run How it works - Install with npx skills add Aidan945/SafeAgentDB --skill safeagentdb, then ask your agent to use the safeagentdb skill in your project - The templates are built for Supabase, Vercel and GitHub Actions; on other stacks the agent proposes an adapted plan before editing - It bundles scripts, GitHub Actions workflows, package scripts and agent instruction files Good to know - It asks before using or setting credentials (a Supabase access token, Vercel token and GitHub secrets) and never commits secrets - It won't copy production data into previews without your explicit approval - Supabase database branching is a paid feature billed per branch-hour; the skill checks costs with you first - MIT licensed