ci-cd-and-automation

Create a workflow file at .github/workflows/ci.yml that triggers on pull requests and pushes to main. Include jobs for lint, type checking with tsc --noEmit, unit tests, build, and npm audit, using actions/setup-node with dependency caching.

Define a Postgres service container in the workflow job with health checks, then run migrations and tests against it using a DATABASE_URL environment variable. Store credentials in GitHub Secrets rather than hardcoding them, even for CI-only databases.

A complete pipeline runs lint, type checking, unit tests, build, integration tests, optional E2E tests, security audit, and bundle size checks in sequence. No gate should be skipped; failures must block merge through branch protection rules.

Slow pipelines usually lack dependency caching and run jobs sequentially. Apply caching via setup-node, split lint/typecheck/test into parallel jobs, use path filters to skip unrelated work, shard tests across runners, and move slow tests off the critical path.

Use feature flags when you want to merge incomplete or risky features to main without enabling them for users. Flags allow canary rollouts, instant rollback without redeploying, and A/B testing, but should be removed after full rollout to avoid technical debt.

Yes, copy the specific CI failure output and give it to the agent with instructions to fix and verify locally before pushing. Lint failures map to auto-fix commands, type errors to targeted fixes, and test failures to a debugging workflow.