What problem does it solve? Tests that pass in isolation but fail in combined pytest sessions are usually victims of shared global state—module-level connection pools, user caches, rate limiters, and CSRF tokens that leak between tests. This Skill documents the observed pollution paths in this repository and provides concrete fixture patterns to make tests reliable in any execution order. ## Core Features & Use Cases - Pollution Path Catalog: Documents four concrete leak sources—_pool_cache in app/models/database.py, _ACTIVE_USER_CACHE in app/api/deps.py, the limiter singleton, and CSRF token state. - Mitigation Patterns: Provides four ready-to-use patterns including autouse reset fixtures, per-test pool creation with dependency overrides, function-scoped fixtures, and rate limiter resets. - Diagnostic Workflow: Gives a step-by-step procedure for reproducing order-dependent failures by running tests in isolation versus combined sessions and grepping for shared globals. - Use Case: When test_vaults.py passes alone but fails in the full suite, use the autouse fixture pattern to clear _pool_cache and _ACTIVE_USER_CACHE between tests. ## Quick Start Ask the assistant to diagnose why a test passes alone but fails in the full pytest suite and recommend an isolation fixture.
