From supervibe
Use BEFORE designing tests for a new feature TO establish pyramid (unit/integration/e2e ratios), fixture isolation policy, flake budget, coverage triangulation strategy. Triggers: 'test strategy', 'покрытие тестами', 'тестовая пирамида', 'стратегия тестов'.
npx claudepluginhub vtrka/supervibe --plugin supervibeThis skill is limited to using the following tools:
BEFORE writing the first test for a new feature. BEFORE adding a new test layer (e.g. introducing Playwright into a project that only had Jest). WHEN flake rate breaches the project SLO. WHEN a coverage report shows numerically high coverage but bugs still ship.
Mandates invoking relevant skills via tools before any response in coding sessions. Covers access, priorities, and adaptations for Claude Code, Copilot CLI, Gemini CLI.
Share bugs, ideas, or general feedback.
BEFORE writing the first test for a new feature. BEFORE adding a new test layer (e.g. introducing Playwright into a project that only had Jest). WHEN flake rate breaches the project SLO. WHEN a coverage report shows numerically high coverage but bugs still ship.
The skill answers four questions in order: what shape is the pyramid, how do fixtures stay isolated, what flake budget is acceptable, how is coverage triangulated.
Follow docs/references/skill-expert-operating-standard.md: start from source of truth, preserve retrieval evidence, apply scope safety, use real producers with runtime receipts for durable delegated outputs, verify before completion claims, and keep confidence below gate when evidence is partial.
package.json / pyproject.toml / Cargo.toml to enumerate which test runners exist already..github/workflows/, .gitlab-ci.yml, etc.) and capture current test job timings.skip, xit, @flaky, t.Skip, quarantine tags).Is the feature pure logic (no I/O, no UI)?
YES → 90% unit, 10% contract; no e2e needed
NO → continue
Does the feature cross a process boundary (DB, HTTP, queue)?
YES → add integration layer with real dependency in container or in-memory double
NO → unit + a single happy-path e2e is enough
Is there a user-visible flow (form → server → screen)?
YES → at least one e2e per critical journey, capped at 10% of total runtime
NO → skip e2e
Is the feature behind a flag with staged rollout?
YES → add a contract test pinned to the OFF state so cleanup is safe later
NO → standard pyramid only
docs/testing.md or equivalent.makeUser({ overrides })) over fixture files. Fixture files are allowed only for golden-file assertions..skip indefinitely.supervibe:confidence-scoring with artifact-type=agent-output; ≥9 required to mark this skill complete.A strategy document containing:
Pyramid: <unit%> / <integration%> / <e2e%> (count + runtime)
Isolation: <transactional | per-test-db | in-memory>
Factories: <path to factory module> + ban list for shared fixtures
Flake budget: <X% over Y days>; quarantine TTL: <Z days>
Coverage triangulation: line ≥ <a>%, branch ≥ <b>%, mutation ≥ <c>% on <modules>
CI gates: <list of failing thresholds>
Fast-feedback: unit suite target = <seconds>; runner = <name>
Open risks: <enumerated>
sleep(500) instead of waiting for a condition; flaky and slow simultaneously..skip with no expiry; signals decay until the suite is ignored.supervibe:tdd — drives the day-to-day red/green/refactor loop on top of this strategy.supervibe:audit — periodic re-check of flake rate and coverage triangulation.supervibe:feature-flag-rollout — paired strategy when the feature ships behind a flag.supervibe:error-envelope-design — defines the contract that integration tests assert against.