AgentSkillsCN

testing-patterns

TDD工作流、测试金字塔、覆盖率、模拟反模式及真理三元组整合。使用渐进式披露，延迟加载语言特定框架语法。

SKILL.md
--- frontmatter
name: testing-patterns
description: TDD workflow, test pyramid, coverage, mocking anti-patterns, and Triad of Truth integration. Uses Progressive Disclosure to lazy-load language-specific framework syntax.
license: MIT
metadata:
  author: groupzer0
  version: "2.0"

Testing Patterns & Methodology

Systematic approach to effective testing. Use this skill when writing tests, designing test strategies, reviewing coverage, or evaluating pull requests.

1. The Triad of Truth (Testing)

Every test strategy or QA review must be synchronized:

  1. •Markdown (agent-output/qa/): Save the detailed test strategy, coverage reports, or QA gaps here.
  2. •Obsidian Graph (workflows/): Create or patch the WF- node linking your QA verdict to the Implementation.
  3. •Planka Board: Check off QA/Testing tasks, manage visual labels (e.g., Tests Passed, Coverage Gap), and add your verdict comment.

2. Tooling Contract (Native MCP & Standard Commands)

CRITICAL RULE: You MUST use standard framework commands and native MCP tools for all testing operations.

  • •To run tests: Execute standard framework commands (npm test, pytest, go test ./...) via the terminal or task runner.
  • •To check coverage: Use the native framework tools (e.g., jest --coverage, pytest --cov).
  • •To validate missing tests: Use the native search MCP tool to cross-reference source files with their *.test.ts or test_*.py counterparts.

3. Test-Driven Development (TDD)

TDD is MANDATORY for new feature code. Write tests before implementation.

The TDD Cycle

  1. •RED: Write a failing test.
  2. •GREEN: Write the minimal code to pass.
  3. •REFACTOR: Clean up, keeping tests green.

If an implementation arrives without tests:

  1. •Reject it with a "TDD Required" verdict.
  2. •Specify which tests should exist.
  3. •Send it back to the Implementer.

4. The Iron Laws of Mocking & Anti-Patterns

Mocks are tools to isolate, not things to test.

The Iron Laws:

  1. •NEVER test mock behavior: Use mocks to isolate your unit from dependencies, but assert on the unit's behavior.
  2. •NEVER mock without understanding: Know the side effects before isolating.
  3. •NEVER create incomplete mocks: Mirror real API structures completely.

Anti-Pattern 1: Testing Mock Behavior

  • •Violation: Asserting that a mock was called or rendered (expect(getByTestId('mock')).toBeInTheDocument()) instead of checking the component's actual output.
  • •Fix: Test real component behavior. Only mock to isolate external boundaries (like network or DB).

Anti-Pattern 2: Test-Only Methods in Production

  • •Violation: Adding a destroy() or reset() method to a production class just so tests can clean up.
  • •Fix: Keep production classes pure. Put test-cleanup logic in test utilities.

Anti-Pattern 3: Mocking Without Understanding & Incomplete Mocks

  • •Violation: Over-mocking to "be safe" or returning only partial objects (e.g., just { id, name }) when the real API returns 20 fields.
  • •Fix: Mock at the correct, lowest level. Mirror the real API completely. Downstream code will fail silently if you hide structural assumptions.

5. Test Frameworks (Progressive Disclosure)

When you need specific syntax or setup boilerplate for a testing framework, use filesystem/read_text_file to lazy-load the reference guide:

Read: .github/skills/testing-patterns/references/testing-frameworks.md

This reference includes detailed syntax for:

  • •JavaScript/TypeScript (Jest, Vitest)
  • •Python (pytest)
  • •Java (JUnit 5, Mockito)
  • •Go (testing package)
  • •VS Code Extensions

6. Coverage & Edge Case Strategy

What to Cover (Priority Order):

  1. •Business logic (revenue-impacting calculations)
  2. •Security boundaries (auth, access control)
  3. •Error paths (exceptions, network failures)
  4. •Integration points (API contracts)

Edge Case Generation Matrix:

  • •Numbers: 0, -1, MAX_INT, floating point precision.
  • •Strings: "" (empty), very long strings, special chars/unicode, whitespace only.
  • •Collections: Empty, single item, 1000+ items, null/undefined elements.
  • •Dates: Leap years, timezone boundaries, past/future extremes.

7. Agent Responsibilities & The Triad Handoff

Before concluding your turn as QA or Implementer:

  1. •The Artifact (agent-output/): Save test results or strategies using filesystem/write_file.
  2. •The Execution (Planka): Check off completed testing tasks and use add_label_to_card (e.g., QA Passed).
  3. •The Memory (Obsidian): Use obsidian/write_note to create your WF-<concrete-id> node (Strictly 10-Line Rule, type: QA or Implementation).

Final Chat Message: Always conclude your turn with:

"Handoff Ready. Parent Node context for the next agent is [[WF-...]] (Planka Card: CARD_ID_NUMERIC)."