mk-qa-manual
Spec-driven manual QA and Playwright E2E generation: navigates apps like a human tester, producing reports or .spec.ts files. Prompts for credentials. NOT for autonomous flows (mk:agent-browser).
- License 许可
- MIT
- 最近更新
- 2026-11-15
- Stars
- ★ 16
- Forks
- 3
01 前言 Front Matter
- name
- mk-qa-manual
- description
- Spec-driven manual QA and Playwright E2E generation: navigates apps like a human tester, producing reports or .spec.ts files. Prompts for credentials. NOT for autonomous flows (mk:agent-browser).
02 资源文件 6
- 参考文档examples/01-login-flow-manual-qa.md1.5 KB
- 参考文档examples/02-checkout-e2e-generation.md5.6 KB
- 参考文档examples/03-url-exploration.md1.6 KB
- 参考文档references/output-templates.md10.4 KB
- 参考文档references/qa-process.md5.1 KB
- 脚本scripts/run-playwright.sh848 B
03 SKILL.md 正文
<!-- Design decisions:
- Dynamic routing model: each action uses the best skill (Phase 1.2 routing tree)
- Auth protocol: prompt-always — NEVER guess credentials (absolute constraint)
- playwright-cli preferred for DOM interaction — generates PW code as side-effect
- agent-browser preferred for auth flows — session persistence support
- Role-based locators mandatory (getByRole, getByLabel) — not CSS selectors
- Script for Playwright execution — deterministic, not LLM judgment -->
QA Manual Testing & E2E Code Generation
Orchestrates browser skills to act like a human tester: navigate, interact, verify, and report.
When to Use
Trigger: "test this flow", "QA this page", "generate E2E tests", "test login flow", "manual testing", "write Playwright tests for"
Explicit:
the qa-manual skill tasks/plans/260315-auth-flow.md --report→ manual QA reportthe qa-manual skill tasks/plans/260315-checkout.md --generate→ Playwright .spec.tsthe qa-manual skill https://app.example.com→ explore and test from URLthe qa-manual skill [spec]→ auto-detect use case from spec content
Workflow Integration
Operates in Phase 4 (Review) for QA reports or Phase 3 (Build) for E2E code generation.
Skill Routing
Each action uses the best browser skill:
| Action | Skill | Why |
|---|---|---|
| DOM interaction (click, fill, type) | mk:playwright-cli |
Generates Playwright code as side-effect |
| Navigation + page load | mk:playwright-cli |
goto + wait, code output |
| Visual verification (screenshot) | mk:agent-browser |
Annotated snapshots |
| Auth flow (login, session, MFA) | mk:agent-browser |
Session persistence, auth import |
| Execute Playwright test | Script: npx playwright test |
Deterministic |
Authentication Protocol
On auth encounter (login page, MFA, 401/403):
- STOP current flow immediately
- Show user:
⚠️ Authentication required at: [current URL] mk:qa-manual needs credentials to continue. Please provide: - Username/email: ___________ - Password: ___________ Type 'skip' to skip this flow or 'abort' to stop testing. - WAIT for user input — do NOT proceed without it
- On credentials → fill form using
mk:playwright-cli, continue flow - On 'skip' → mark flow SKIPPED, move to next
- On 'abort' → end session, output partial report
NEVER: Guess credentials. Use hardcoded test credentials. Proceed past auth without input. Store credentials in any file.
Process
See references/qa-process.md for detailed Use Case A and Use Case B processes.
Use Case A — Manual QA: Load spec → extract flows → navigate each → verify outcomes → report
Use Case B — E2E Code Gen: Run Use Case A → record interactions → generate .spec.ts with role-based locators
Output Format
See references/output-templates.md for full templates.
QA Report → structured pass/fail per flow with evidence
Playwright Code → production-ready .spec.ts using getByRole, getByLabel, getByText
Scripts
| Script | Purpose | Invocation |
|---|---|---|
scripts/run-playwright.sh |
Execute generated .spec.ts | After E2E code generation |
References
| Reference | When to load | Content |
|---|---|---|
| qa-process.md | Steps 1-4 | Detailed Use Case A + B processes |
| output-templates.md | Step output | QA report + Playwright code templates |
Failure Handling
| Failure | Detection | Recovery | Message |
|---|---|---|---|
| Auth encountered | URL/form patterns | Auth protocol above | "⚠️ Authentication required..." |
| Page load timeout | 30s timeout | Retry ×2 then report | "Page load timed out at [URL]" |
| Element not found | Selector fails | Re-snapshot, try alternate | "Element not found, re-scanning..." |
| Spec not found | File missing | Ask user | "Spec file not found. Provide path or URL." |
| Playwright exec fails | Non-zero exit | Show error + fix hints | "Test failed: [error]. Fix: [hint]" |
Gotchas
- Testing against stale deployment: Running E2E tests against yesterday's build → Verify deployment version matches expected before starting test run
- Credentials hardcoded in generated spec files: Auto-generated .spec.ts contains login credentials → Always use environment variables for credentials; never inline in test code
04 同仓库更多 Skills
mk-worktree
Manages git worktrees for parallel agent isolation: creates parallel/{name}-{timestamp} branches, merges with conflict detection. NOT for user feature branches (use git worktree directly).
mk-workflow-orchestrator
Auto-invoked 7-phase workflow for complex-feature intent; fasttrack mode for pre-approved specs. NOT for explicit single tasks (mk:cook), green-field builds (mk:autobuild), or simple fixes (mk:fix).
mk-web-to-markdown
Fetch a URL and return clean markdown, for arbitrary external pages (blog, RFC, GitHub issue, vendor doc) not covered by mk:docs-finder. Static-only by default; opt-in JS rendering available.
mk-vulnerability-scanner
Vulnerability analysis: OWASP Top 10, current threat landscape, supply chain security, attack surface mapping, risk prioritization. Use for security audits. NOT for STRIDE/CSO infra audits (mk:cso).
mk-vue
Vue 3 authoring/review: components, composables, reactivity, Pinia, Pinia Colada, file routing, forms. Composition API + script setup. Auto-activates on .vue files. Deep review: mk:vue-best-practices.
mk-vue-testing-best-practices
Vue 3 testing best-practices review: Vitest + Vue Test Utils for components, composables, Pinia, Router, async/Suspense, forms, a11y, plus Playwright E2E strategy. Advisory only.