Technical Testing Posture 🧪🏗️

"Tests are not documentation; they are the executable guardrails of the architecture."

This addon is triggered for any task involving Unit, Integration, or E2E tests, CI verification, or Coverage auditing.

When to use

Load this skill when:

• You are writing new tests (Vitest, Playwright).
• The task involves fixing a bug (Test-First required).
• You are optimizing the CI test pipeline.
• You are auditing test coverage for a mission-critical module.
• The work involves mocking, spying, or environment orchestration.

Operating Rules (R1–R5)

R1 - Integration-First Philosophy (V1, S1, S5)

Unit tests are for logic; Integration tests are for reality. Prioritize testing the boundaries between modules (e.g., Service -> DB, Controller -> Service). A passing unit test for a broken SQL query is a failure of posture.

R2 - Production-Parity (V20, S20)

Integration and E2E tests MUST run against real infrastructure clones (e.g., Docker-based PostgreSQL, Redis) instead of mocks. Mocks are for 3rd-party APIs only. If the DB behavior differs between test and prod, the test is invalid.

R3 - Mandatory Edge-Case Coverage (V7, S6, S14)

The "Happy Path" is the baseline. 80% of test logic should focus on the "Unfiltered Reality": network timeouts, database deadlocks, validation failures, and unauthorized attempts. If a function can throw, there must be a test for it.

R4 - Test-First Restoration (V8, S21)

Every bug fix MUST start with a failing test that reproduces the issue. This prevents regressions and establishes the "Ground Truth" of the fix. Never submit a fix without its reproductive test.

R5 - Flake-Zero Resilience (V4, S1)

A flaky test is worse than no test. Avoid sleep() or arbitrary timeouts; use polling and "Eventually" assertions. Tests must be isolated and capable of running in parallel without side effects on shared state.

Testing Self-Verification Gate

• Boundary: Does this test a real integration (e.g., DB call) or just a mock?
• Parity: Is the test running against a production-mirror environment?
• Edges: Are there tests for at least 3 failure scenarios (timeout, validation, 403)?
• Reproduction: If this is a fix, is there a test that failed before the code change?
• Resilience: Are there any non-deterministic sleep calls?

Deep Reference Library

• references/00-dossier.md - The Principal Constitution (V1-V36, S1-S41).
• references/06-the-canon.md - The Intellectual Compass (Testing theory).
• references/08-principal-mindset.md - Confidence-based rule precedence.
• references/09-calibration-diffs.md - 12 high-density Before vs. After examples.