Goal-Loop methodology for multi-step agent work: strategy exploration, GOAL/GOALS/REPORT files, sub-goal state machine, iteration budgets, rollback, and decision logs. Use when the user asks for Goal-Loop, Goal Kit, GOAL.md, GOALS.md, REPORT.md, .hermes planning, multi-phase delivery with checkpoints, or structured goal decomposition before execution.
Use for rule-engine and domain logic tests—VX-01~12 cross-consistency, W-EPC-01~17 linter, epc-coverage computeCoverage, deriveEpcSteps ordering. Table-driven; one positive and one negative case per rule minimum.
Use for tests/e2e smoke flows—ModelingWorkspace, business chain navigation, C workspace panels, derive/apply EPC. Tag @smoke; use waitFor; never mix vi.useFakeTimers with waitFor.
Use for tests/integration—React components with @testing-library/react, ScenarioWorkspace, panels, element-library. Focus on testids, tabs, empty states, and store wiring without full E2E.
Use before Coding (Unit step 3)—translate Unit Spec AC into Testing Cases with file paths, TC IDs, and pass/fail scenarios. Must complete before any production code for the Unit.
Use when orchestrating a US/Unit test effort—assigning subagents, defining exit criteria, and signing off only with fresh verification evidence. Test Lead does not write all tests; it plans, dispatches, and gates.
Use for tests/unit and store integration—pure functions (lib/*), Zustand store getters, and helpers. RED-GREEN with vitest; no React unless unavoidable.