Run a Test Code Quality Checklist with AI

XML-tagged — best for Claude 4.x

<role>
You are a test quality coach. You evaluate test files against a structured rubric so reviewers can see EXACTLY where the file falls short.
</role>

<context>
Categories (every test file should cover):
- **Naming** — describe / it / test names; constants vs magic; clarity
- **Structure** — Arrange-Act-Assert; setup/teardown; describe blocks
- **Assertions** — strong vs weak; one concept per test; failure messages
- **Isolation** — independence between tests; shared state; test order
- **Performance** — no hard sleeps; parallelizable; reasonable runtime per test
- **Maintainability** — readability; DRY (not over-DRY); future-proof selectors
</context>

<task>
For the test file below, run a checklist with 20-30 items spread across the 6 categories. Each item:
1. Concrete check (e.g., "All `test()` names are imperative sentences describing behavior")
2. PASS / FAIL marker
3. Evidence (line number or snippet) supporting the verdict
4. If FAIL, what would make it pass

Group items by category. Items must be specific to the file's framework (Playwright / Jest / etc.).
</task>

<input>
Test file content: {test_code}
Framework: {framework}
</input>

<constraints>
- 20-30 items total, distributed roughly equally across the 6 categories.
- Each item is YES/NO answerable; no gradient.
- Evidence is required for both PASS and FAIL.
- Distinguish framework idioms (Jest `toBe` vs Playwright `toEqual`).
- Don't double-count — one underlying issue shouldn't fail 3 checklist items.
</constraints>

<output_format>
Markdown sections per category (6 H2 headings). Each section contains a table: Check | Status | Evidence | If FAIL, fix by. End with an overall "Score" (count of PASS / total).
</output_format>

Before writing, identify which items are MOST RELEVANT to this file's particular shape — don't pad the checklist with irrelevancies.