Test Strategy
| Title: Test Strategy |
Version: 1.0 |
Owner: [TENANT_CONFIGURATION_REQUIRED — QA Lead] |
Status: Draft |
Last reviewed: 2026-09-07 |
Next review: [TENANT_CONFIGURATION_REQUIRED] |
Reviewers: QA, Architecture, Security, AI Governance |
Purpose and scope
Defines the full test pyramid for the platform, integrating functional, security, performance, accessibility, and AI-specific testing tiers.
Test pyramid
flowchart TB
A[Unit tests] --> B[Integration tests]
B --> C[Contract tests - OpenAPI/AsyncAPI]
C --> D[End-to-end tests]
D --> E[Security tests]
E --> F[Performance tests]
F --> G[Accessibility tests]
G --> H[Chaos/resilience tests]
H --> I[RAG evaluation]
I --> J[Agent/skill evaluation]
J --> K[MCP integration tests]
K --> L[Human acceptance testing]
Tier definitions
| Tier |
Scope |
Owner |
Cadence |
| Unit |
Individual functions/classes, domain invariants (e.g., illegal state transitions rejected) |
Engineering |
Every commit |
| Integration |
Service-to-DB, service-to-service within a bounded context |
Engineering |
Every PR |
| Contract |
Requests/responses conform to hr-onboarding-api.openapi.yaml / hr-onboarding-events.asyncapi.yaml |
Engineering |
Every PR |
| End-to-end |
Full workflow paths (see test-case-catalog.md) |
QA |
Every release candidate |
| Security |
Per security-test-plan.md |
Security |
Every PR (SAST/SCA) + every release (DAST/pen-test cadence) |
| Performance |
Load/latency against non-functional-requirements.md targets |
Performance Engineering |
Every release, plus scheduled load tests |
| Accessibility |
WCAG target conformance |
QA/Frontend |
Every UI-affecting release |
| Chaos/resilience |
Fault injection per resilience-and-disaster-recovery.md |
SRE |
Scheduled cadence in staging |
| RAG evaluation |
Groundedness, citation validity, no-answer calibration |
AI Governance |
Every prompt/model/RAG-config change |
| Agent/skill evaluation |
Functional accuracy, fairness, per ai-evaluation-strategy.md |
AI Governance |
Every prompt/model change |
| MCP integration |
Tool schema conformance, approval-gating enforcement, circuit breaker behavior |
Integration Engineering |
Every PR touching adapters/MCP |
| Human acceptance |
Business stakeholder sign-off against acceptance-criteria.md |
Product/HR |
Every release |
Test data policy
All test data is fake/demo — never real candidate/employee data (Section 7, CLAUDE.md). Synthetic data generation for volume/load testing must not be derived from de-anonymized production data.
Change control
| Version |
Date |
Author |
Change |
| 1.0 |
2026-09-07 |
Documentation package generation |
Initial creation |