菜单
Test code quality review for Playwright specs. Deterministic regex catalog in scripts/test_quality_rules.json scanned by hooks/test-quality-scan.ps1; agent only does holistic judgments (assertion semantics, locator-strategy fit, isolation).
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。
基于 SOC 职业分类
Azure DevOps work-item and PR operations — read work items, create new bugs/tasks, resolve work items, create PRs, and full automated accessibility fix loops. Uses az CLI with the azure-devops extension, with MCP fallback (mcp__dynamicscrm-repo__* or mcp__msazure-repo__*). Reads repo configuration from ~/.pwagent/config.json.
Project-specific S360 fix patterns for pwagent. Read by pwagent-s360 --fix before writing any code. Contains Kusto query patterns, portal URL templates, ADO field mappings, and per-subtype remediation notes specific to this workspace.
Battle-tested Playwright patterns for writing, debugging, and scaling reliable test suites. E2E, API, visual, accessibility, security testing, plus CI/CD, CLI automation, and page objects. TypeScript and JavaScript.
Run Kusto Query Language (KQL) queries to read data from Azure Data Explorer / Geneva / pipeline telemetry clusters. Read-only — never mutates state. Agents call this to fetch flake history, test run summaries, telemetry signals.
Production-ready CI/CD configurations for Playwright — GitHub Actions, GitLab CI, CircleCI, Azure DevOps, Jenkins, Docker, parallel sharding, reporting, code coverage, and global setup/teardown.
Battle-tested Playwright patterns for writing and debugging reliable E2E, API, component, visual, accessibility, and security tests. Use when you need locator strategy, assertions, fixtures, network mocking, auth flows, trace debugging, or framework recipes for React, Next.js, Vue, and Angular. TypeScript and JavaScript.
| name | test-review |
| description | Test code quality review for Playwright specs. Deterministic regex catalog in scripts/test_quality_rules.json scanned by hooks/test-quality-scan.ps1; agent only does holistic judgments (assertion semantics, locator-strategy fit, isolation). |
| allowed-tools | Read Bash(pwsh:*) |
Per N2 (SPEC.md §16) the deterministic part of test review (regex against the catalog) runs as hooks/test-quality-scan.ps1. The agent reads the resulting sidecar + adds holistic findings.
The rule catalog lives at scripts/test_quality_rules.json. Each rule:
{
"id": "PW-001",
"severity": "High|Medium|Low",
"pattern": "<regex matched against file content line>",
"message": "<one-line guidance>",
"fix_guide": "<relative path under skills/>"
}
Current catalog (operationalizes skills/SKILL.md's Forbidden list + Golden Rules):
| ID | Severity | Catches |
|---|---|---|
| PW-001 | High | page.locator('css=...') |
| PW-002 | High | page.locator('xpath=...') or page.locator('//...') |
| PW-003 | High | page.locator('.class') or page.locator('#id') |
| PW-004 | High | page.$(...) or page.$$(...) (handle APIs) |
| PW-005 | High | page.waitForSelector(...) |
| PW-006 | High | page.waitForTimeout(...) |
| PW-007 | Medium | expect(await locator.textContent()) (non-web-first) |
| PW-008 | Medium | Hardcoded https?://... URL |
| PW-009 | Medium | retries: <non-zero-non-2> |
Add new rules by editing scripts/test_quality_rules.json and re-running the scan. No code changes needed.
The agent reads runs/<runId>/.test-quality-findings.json (written by the hook) and asks, per test file:
getByRole replace a getByTestId?beforeEach blocks polluting state?Output ≥3 semantic findings per file (or zero with "no semantic issues" if clean).
| Severity | Action |
|---|---|
High | Always file an ADO bug; block the agent's PR if introduced new (/pwagent-fix integration) |
Medium | PR comment via az repos pr comment create if --pr-comment <id> flag |
Low | Informational; appears in dashboard only |
~/.copilot/state/runs/<runId>/test-review.md:
Every finding cites the literal matched line (from the hook's matched field).