What problem does it solve? Teams with all-mocked unit test suites keep hitting production failures at the seams — SQL constraint violations, serialization mismatches, framework wiring bugs — because mocks only encode beliefs about dependencies, never verify them. This Skill provides a workflow for deciding what to test with real components wired together versus unit tests. ## Core Features & Use Cases - Seam Selection: Identifies where integration tests belong (real SQL execution, ORM mappings, queue serialization, HTTP contracts) and enforces one real integration per test. - Real Engines via Testcontainers: Replaces H2/SQLite/fakeredis lookalikes with pinned-version containers (Postgres, Redis, Kafka, LocalStack) so tests match production behavior. - Structural Isolation & Migration Testing: Covers transaction-rollback or schema-per-worker isolation and building test schemas via real migrations instead of create_all(). - Contract Testing: Uses Pact or recorded fixtures for cross-service boundaries instead of booting other teams' services in CI. - Use Case: A payments service with 400 mocked unit tests suffers repeated SQL constraint and Kafka serialization incidents; applying this Skill adds 25 testcontainer-backed repository tests, 6 serialization round-trip tests, and Pact contracts, eliminating seam incidents the next quarter. ## Quick Start Ask the AI to design an integration test strategy for your service, specifying which dependencies (database, queue, downstream APIs) should be tested with real components versus stubs.