ci-cd-and-automation

Create a workflow file at .github/workflows/ci.yml that triggers on pull requests and pushes to main. Add jobs that install dependencies with npm ci, then run lint, tsc --noEmit, tests with coverage, build, and npm audit in sequence.

Define a Postgres service container in the job with health checks, then run migrations and integration 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 includes lint, type checking, unit tests, build, integration tests, optional E2E tests, security audit, and bundle size checks. No gate should be skipped, and branch protection should block merging until all checks pass.

Slow pipelines usually lack dependency caching and run jobs sequentially. Apply the setup-node cache option, split lint, typecheck, and test into parallel jobs, use path filters to skip unrelated work, and shard large test suites across runners.

Use feature flags when you want to merge incomplete or risky features to main without enabling them for users. Flags let you canary releases to a percentage of users, roll back without redeploying, and should be removed after full rollout to avoid technical debt.

Yes, copy the specific failure output from the CI run 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 the reported location, and test failures to a debugging workflow.