Validates Jest tests pass both as a full suite and in isolated runs.
npx skills add https://github.com/mozilla/fxa --skill fxa-test-independence
Please help me install this Agent Skill. Skill: fxa-test-independence Source: https://github.com/mozilla/fxa/tree/main/.claude/skills/fxa-test-independence Command: npx skills add https://github.com/mozilla/fxa --skill fxa-test-independence
Jest tests that pass in a full suite but fail when run alone hide order dependencies and shared mutable state, masking real bugs in the test suite. This Skill detects those hidden dependencies by running every test twice and comparing results.
it(...)/test(...) individually with --testNamePattern from the correct package root.beforeEach resets or polluted shared mocks.packages/fxa-auth-server/lib/account.spec.ts, invoke the Skill to confirm all 12 tests pass both together and in isolation before committing.Ask the assistant to validate test independence for a given spec file, for example: run fxa-test-independence on packages/fxa-auth-server/lib/metricsCache.spec.ts.
High-intent search queries and answers about installing and using this skill.
Run the full spec file once, then re-run each test individually using npx jest with --testNamePattern for the exact test name. A test that passes in the suite but fails alone has an order dependency on state set by a prior test.
Change into the package root so the local jest.config is picked up, then run npx jest <relative-spec-path> --testNamePattern="<exact test name>" --no-coverage. Avoid nx test-unit, which runs the entire package suite regardless of the file argument.
This indicates an order dependency: a prior test mutates shared module-level state, a singleton is not re-initialized, or a missing beforeEach reset leaves stale data. The fix is to isolate state per test rather than change the implementation.
That is a false positive pattern: the test pollutes shared state, such as mutating a global mock without an afterEach cleanup, which breaks other tests. Add proper teardown so each test leaves shared state untouched.
No. It reports a results table with failure output, a diagnosis of likely causes, and a suggested concrete fix for each non-OK test, but the engineer decides whether and how to apply the fix.
open-mercatoRuns only Jest and Playwright tests affected by changed code in a monorepo.
richardwuFixes GitHub pull request review comments and CI check failures locally, commits and pushes updates.
ArthTech-AIRun full-stack QA checks with pytest, Jest, ruff, TypeScript, and Playwright.
dorrioAudit software test suites for isolation and determinism issues.