e2e-testing
GitHub基于 Playwright 编写、修复和诊断 E2E 测试，复用现有套件与 fixtures，通过角色定位元素并断言可见行为。针对失败情况利用 trace 和截图进行调试，确保测试独立且准确反映用户可见功能。
Trigger Scenarios
Install
npx skills add nodetool-ai/nodetool --skill e2e-testing -g -y
SKILL.md
Frontmatter
{
"name": "e2e-testing",
"origin": "ECC",
"description": "Write, repair, or diagnose Playwright E2E tests for NodeTool web and workflow harnesses using existing suites and fixtures."
}
E2E testing
Build or repair a Playwright test that detects the requested user-visible behavior.
Use web/TESTING.md and the existing suite's fixtures,
page objects, and configuration. Do not replace them with a generic scaffold.
Electron main-process tests use Jest under electron/src/__tests__/, as described
in Electron instructions.
Choose the suite
Inspect web/package.json and the relevant Playwright config for the exact command.
Use the graph runner or debug harness for real-backend workflow scenarios. Read
its entry in the harness reference before first use.
Build backend packages and install the required Playwright browser as described
in the testing guide before running web E2E.
Write or diagnose the test
Reuse the nearest scenario. Select controls by role and accessible name when available. Assert the specific visible result with retrying Playwright assertions. Keep tests independent and reset fixtures through the suite's existing mechanism.
Register a network wait before triggering the action that produces the request. Wait for a specific response or visible state, not arbitrary sleeps or global network idleness. Use page objects only where they simplify repeated interactions.
For a reported failure, reproduce it before changing the test or application. Inspect its trace, screenshot, console, and request evidence. Use bounded repeat runs when diagnosing an intermittent race. A retry reaching green does not prove the race is fixed. Do not hide failures with skip, fixme, or extra retries.
Verify and report
Run the affected scenario and confirm the assertion detects the intended failure and passes with the correction. Use the suite's configured trace, screenshot, and video capture rather than introducing a separate recording mechanism.
Complete mandatory post-change verification for code changes. Report the tested scenario, actual command result, and relevant failure artifacts. Do not run unrelated browser suites without a dependency or unresolved concern that warrants them.
Version History
- 9b979bc Current 2026-09-22 23:26
