deployment-rollback

The Deployment Rollback skill provides a comprehensive operational toolkit for maintaining production stability in Vercel-hosted environments. It is designed for DevOps engineers, site reliability engineers (SREs), and backend developers who need to respond immediately to critical failures, performance regressions, or botched deployment cycles. This skill centralizes emergency recovery procedures, ensuring that restoration processes are repeatable, audit-ready, and safe.

• Executes Vercel-specific rollback commands to promote previous stable deployments or revert to known-good Git commits.

• Performs database-level operations, including rolling back Drizzle ORM migrations or restoring PostgreSQL data from pg_dump backups.

• Manages infrastructure cache clearing, specifically targeting Next.js revalidation endpoints and Redis FLUSHALL operations to resolve stale data issues.

• Implements post-rollback verification protocols using automated health checks for web endpoints and database connectivity diagnostics.

• Provides a structured, checklist-driven approach to incident response, ensuring pre-rollback safety checks and post-rollback documentation are performed.

• Supports Git-based workflow management, such as creating emergency hotfix branches, performing git reverts, and triggering redeployments via PRs.

• Designed for high-pressure production environments where every minute of downtime impacts user experience.

• Leverages the Vercel CLI, Vercel Dashboard tools, and Hono/Next.js infrastructure constraints.

• Requires appropriate authentication tokens (REVALIDATE_TOKEN) and database access credentials (DATABASE_URL).

• Essential for scenarios involving critical bugs identified post-deployment, unexpected performance drops, or database schema mismatches.

• Integration with monitoring tools (e.g., Vercel analytics or custom health checks) is encouraged to validate success before resuming normal operations.

• Follows best practices like utilizing feature flags to bypass full rollbacks when possible and documenting post-mortem outcomes for every emergency action taken.