Detect the runner, environment, Rstack integration, and smallest runnable scope with references/detect-test-framework.md.
Run references/dependency-install-gate.md before choosing Rstest, coverage, adapter, or plugin APIs.
Before editing, record the exact command, Node and runner versions, environment, files/tests/skips/snapshots, and failures. Capture the pre-migration test manifest and follow references/discovery-parity.md throughout the migration.
For Jest read references/jest-migration-deltas.md; for Vitest read references/vitest-migration-deltas.md; for global APIs also read references/global-api-migration.md; for a detected Playwright Test runner read references/playwright-migration-deltas.md. Do not load the Playwright path for a browser-only request whose runner is not Playwright Test. Migrate scripts, config, and setup before editing test bodies; prefer adapter or Rsbuild/Rspack fixes over broad test rewrites.
If config loading emits [MODULE_TYPELESS_PACKAGE_JSON], use references/config-module-type.md. Do not suppress the warning or change the whole package module type without an audit.
Get the migrated scope semantically green. Fix failures in this order: dependency skew, config/resolver, discovery, setup/environment/coverage, mocks/timers/snapshots, then test bodies.
Compare the post-migration manifest and execution counts with the baseline. Classify every added, removed, skipped, or excluded test explicitly; do not call a run equivalent merely because it is green. For Jest snapshot migrations, also complete the snapshot cleanup and unchecked validation in references/jest-migration-deltas.md.
Remove temporary aliases, diagnostic hooks, and copied legacy workarounds that are no longer required. Keep only configuration proven necessary by behavior or measurement.
If wall time, build, test runtime, logs, or memory materially regress, load and follow the rstest-debugging skill when available. Otherwise use the migration fallback in references/performance-diagnosis.md, then read references/dependency-bundling-performance.md and/or references/mocked-module-build-graph.md according to the measured bottleneck. Change one variable at a time.
After correctness and performance validation, remove only legacy files and dependencies owned by the migrated scope. Summarize behavior parity, discovery differences, unsupported fields, retained compatibility config, performance tradeoffs, and TODOs.

Guardrails

Keep the smallest viable scope. Do not broaden a monorepo migration because scopes share a lockfile.
Do not change production behavior, assertions, test names, scenarios, coverage thresholds, or ignore directives to make migration pass.
Do not introduce jest/vi shims or aliases; rewrite call sites with references/global-api-migration.md.
Do not silently drop unknown config fields or historical excludes. Verify, replace, or report each one.
Do not compare performance across different test manifests, Node versions, coverage modes, cache states, or worker settings.
Keep the previous runner until Rstest is green. Use a local Rstest checkout only for labeled diagnostics, then validate the final result with the project's installed dependency.
Escalate before many test edits or any production-source change: report why smaller config/setup fixes failed, options, risks, and the recommended path.

High-risk Rstest deltas

rstest / rstest run is single-run; watch mode is rstest --watch or rstest watch.
globals defaults to false; preserving global APIs requires globals: true and @rstest/core/globals types.
Rstest builds before tests run. rs.mock() is runtime replacement and does not by itself prune the real module graph.
Dependency bundling trades compiler work for runtime Node loading. Neither “bundle all” nor “externalize all” is universally faster; measure a representative file and the full scope.
An externalized runtime mock must match the exact request and module format. A config-wide external is unsafe when any test needs the real module.

migrate-to-rstest

Migrate to Rstest

Goal

Workflow