What problem does it solve? Changing behavior in a multi-tenant platform like EVA risks breaking tenant isolation, approvals, task state transitions, and route access rules. This Skill enforces a disciplined red-green-refactor loop so every behavior change is captured by a failing test before any implementation begins. ## Core Features & Use Cases - Red-Green-Refactor Loop: Write the smallest failing test, confirm it fails for the expected reason, implement the minimal fix, then refactor while keeping tests green. - EVA Test Targets: Unit tests for services, repositories, guards, and state machines; E2E tests for API flows and mocked DB behavior; RLS verification via RLS_TEST=true npm run test:e2e. - Risk-Based Rule of Thumb: Mandates tests whenever changes touch tenant isolation, approvals, secrets, task state, or public/private route behavior. - Use Case: When fixing a bug in a NestJS guard that controls route access, first write a failing unit test reproducing the unauthorized access, then implement the minimal guard fix and run broader E2E tests to confirm no regressions. ## Quick Start Ask the agent to fix the bug in the task state machine using test-driven development, starting with a failing test that reproduces the incorrect transition.
