Imported from kanhaya/playwriteAgents (
.github/agents/AGENTS.md). Install upstream withnpx skills add kanhaya/playwriteAgents --skill agents. Copyright stays with the author.
Shared Agent Context
All agents in this repository must follow these conventions. Read this file before planning, generating, healing, or reviewing tests.
Application under test
- URL:
https://rahulshettyacademy.com/seleniumPractise/#/ - Type: E-commerce demo (Selenium Practice Store)
- Routing: Hash-based (
#/); URL assertions should allow optional hash fragment - Promo codes: Site may support promo/discount flows — verify in UI before asserting
Repository layout
| Path | Purpose |
|---|---|
specs/test.plan.md |
Master test plan (human + machine-readable YAML at bottom) |
specs/archive/ |
Versioned plan history |
specs/flaky-registry.md |
Quarantined or repeatedly healed tests |
specs/run-reports/ |
Agent run summaries |
tests/*.spec.ts |
Playwright test specs |
tests/pages/ |
Page Object Model classes |
tests/data/ |
JSON test data |
tests/seed.spec.ts |
Seed template for generator |
utilities/ |
Shared helpers (a11y, resolveCtor) |
config/ |
Environment config |
playwright.config.ts |
Playwright runner config |
Environment
TEST_ENV—dev(default),qa,staging,prodBASE_URL— set per environment inplaywright.config.ts- Run CI locally:
npm run test:ci - Start MCP server:
npm run mcp
Locator strategy (priority order)
page.getByRole()with accessible namepage.getByTestId()/data-testidpage.getByLabel()/page.getByPlaceholder()page.getByText()with exact or regex- CSS selectors — only when no semantic locator exists
Page Object Model
- Extend
BasePagefor shared navigation/waits - Existing pages:
HomePage,CartPage,CheckoutPage,ProductPage - Reuse POM methods — do not duplicate selectors in specs
Spec file contract
Every spec must start with:
// spec: specs/test.plan.md
// seed: tests/seed.spec.ts
// generated-by: playwright-test-generator | manual
// generated-at: YYYY-MM-DDHealer edits must add:
// healed-by: playwright-test-healer
// healed-at: YYYY-MM-DD
// reason: <brief explanation>Never do
- Do not use
waitForLoadState('networkidle')— deprecated and flaky - Do not use hardcoded
page.waitForTimeout()— useexpectorlocator.waitFor - Do not hardcode credentials or secrets
- Do not edit
.env*files - Do not modify application source code — only
tests/,specs/,utilities/,scripts/
Agent pipeline
Planner → specs/test.plan.md
Generator → tests/*.spec.ts
Runner → npm run test:ci
Healer → fix failing specs (max 3 iterations per test)
Reviewer → PR quality checklist
Agents available
| Agent | File | When to use |
|---|---|---|
| Orchestrator | playwright-test-orchestrator.agent.md |
Full plan → generate → run → heal loop |
| Planner | playwright-test-planner.agent.md |
Explore app and write test plan |
| Generator | playwright-test-generator.agent.md |
Turn plan scenarios into Playwright specs |
| Healer | playwright-test-healer.agent.md |
Debug and fix failing tests |
| Reviewer | playwright-test-reviewer.agent.md |
Review test quality on PRs |
| Architect | playwright-framework-architect.agent.md |
Framework design and structure |
Invoking agents
# Copilot CLI
copilot --agent playwright-test-orchestrator --prompt "Cover checkout flows end to end"
# VS Code / Cursor: select agent from Agents panel, then promptCI expectations
- PRs: smoke suite + plan drift check
- Nightly: full regression across chromium + firefox
- On nightly failure: healer workflow opens fix PR or documents skip