菜单
Write tests with TDD following structured patterns. Ensures consistent AAA structure, proper coverage targets, and framework-specific conventions. Without this skill, tests lack consistent naming, miss coverage targets, and skip anti-pattern checks.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/nguyenthienthanh/aura-frog --skill test-writer命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Vision feedback loop for UI — render the running UI, screenshot it across viewports + dark mode, run deterministic gates then a multimodal critique against the design system, and iterate until it passes. Use after building or changing a UI component/page, when asked to 'check the UI', make it 'look right', or verify a build matches a design spec / Stitch screen.
UI/UX design expertise — component design, design system selection, responsive layout. Includes auto-detection from package.json and Context7 integration for library docs.
Generate a cohesive color-token system from ONE OKLCH brand hue — primary/secondary/background/foreground/muted/border with semantic light + dark, emitted as a Tailwind v4 @theme block or a CSS-variables fallback. Use when starting a design system, defining brand colors, setting up theming/dark-mode, or replacing scattered hardcoded colors with a derived palette.
基于 SOC 职业分类
正在显示 SKILL.md
| name | test-writer |
| description | Write tests with TDD following structured patterns. Ensures consistent AAA structure, proper coverage targets, and framework-specific conventions. Without this skill, tests lack consistent naming, miss coverage targets, and skip anti-pattern checks. |
| autoInvoke | true |
| priority | medium |
| triggers | ["add tests","test coverage","write tests","TDD","E2E test"] |
| allowed-tools | Read, Grep, Glob, Edit, Write, Bash |
| user-invocable | false |
AI-consumed reference. Optimized for Claude to read during execution. Human-readable explanation: see docs/architecture/HIERARCHICAL_PLANNING.md or docs/getting-started/ depending on topic.
Adding tests, improving coverage, TDD (Phase 2/3). NOT for: bug fixes (use the bugfix-quick skill via the Skill tool — it is NOT an agent), features (run-orchestrator skill).
Test Type Selection below — unit, integration, or e2e (NOT default to unit)npx playwright test printed pass/fail per spec, not just "0 tests found")The default in v3.7.0 → v3.7.3 was implicitly "unit test only" because the framework table didn't list Playwright and no rule prescribed e2e. v3.7.4 fixes this:
test_pyramid[3]{level,trigger,framework,when_required}:
unit,"Pure functions · business logic · isolated components","jest|vitest|pytest|phpunit|go test","always — fast feedback, mocks externals"
integration,"Multi-module interaction · DB queries · API contracts","Same runner + real DB/fakes","whenever ≥ 2 modules cross-talk"
e2e,"User flows · login/checkout/critical paths · cross-page nav","playwright|cypress|detox","UI/auth/payment task OR Phase 2 contract mentions 'flow' / 'journey' / 'end-to-end'"
Trigger words in the task description that REQUIRE an e2e layer in Phase 2:
login, signup, checkout, payment — critical user flows must have at least one e2e specflow, journey, end-to-end, happy path, smoke testacceptance references user-visible behaviorWhen in doubt for a UI task, write BOTH: unit tests for the component logic + at least one e2e spec for the happy path. Skipping the e2e layer is the silent failure mode of v3.7.0–v3.7.3 and is exactly what this section exists to prevent.
If the project has no e2e runner configured yet:
playwright.config.*, cypress.config.*, .detoxrc.js, e2e/ directory.install playwright, use cypress, or unit-only to proceed."principles[5]{principle}:
AAA pattern: Arrange → Act → Assert
"should [action] when [condition]" naming
One concern per test — no shared state
Test behavior not implementation
Mock external deps only — use fakes for complex deps
coverage[4]{scope,target}:
Critical paths (auth/payment),100%
Business logic,90%
UI / utilities,80%
Overall minimum,80%
avoid[6]{pattern,fix}:
Test implementation details,Test behavior/output only
Shared state between tests,Reset in beforeEach
Sleep/delays,Use waitFor/polling
Giant multi-assertion tests,One concern per test
No assertions,Always assert expected outcomes
Unit-only for UI/auth/payment tasks,"Add an e2e layer (playwright) — silent v3.7.0–v3.7.3 default fixed in v3.7.4"
frameworks[8]{framework,layer,file_pattern,runner,config}:
Jest/Vitest,unit,"*.test.ts *.spec.ts","npm test / npx vitest","jest.config.* / vitest.config.*"
PHPUnit,unit,"*Test.php","./vendor/bin/phpunit","phpunit.xml"
Pytest,unit,"test_*.py","pytest --cov=.","pytest.ini / pyproject.toml [tool.pytest]"
Go,unit,"*_test.go","go test -cover ./...","go.mod"
Playwright,e2e,"*.spec.ts (inside tests/ or e2e/)","npx playwright test","playwright.config.*"
Cypress,e2e,"*.cy.ts (inside cypress/e2e/)","npx cypress run","cypress.config.*"
Detox,e2e (mobile),"*.e2e.ts","detox test",".detoxrc.js"
Vitest browser mode,integration,"*.browser.test.ts","npx vitest --browser","vitest.config.* with browser block"
Playwright is the v3.7.4+ recommended e2e default — it ships an MCP server (playwright) on the tester agent's allowlist, supports all 3 major browsers, and integrates with the vitest MCP for one-tool test execution. Cypress and Detox remain supported.
Runner verification — never trust silence. After npx playwright test / npx cypress run, read the actual output: did it list specs, did pass/fail counts appear, is the exit code 0? "Tests passed" without seeing pass count = 0 tests collected = bug. Same discipline applies to all runners.
rules/core/tdd-workflow.md — RED → GREEN → REFACTORrules/core/verification.md — Read output, then claimrules/core/code-quality.md — Coverage targets per file typerules/workflow/post-implementation-linting.md — Lint after tests addedrules/core/simplicity-over-complexity.md — One concern per test, no shared mutable state, no over-engineered mocks