Content
81%Weight 40%Scale 1-5Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A strong, highly actionable skill body: complete executable setup and run commands, a sequenced workflow with a real failure-feedback loop, and well-organized external templates. Its main weakness is token efficiency — the locator and assertion sections re-teach Playwright basics Claude already knows, and a couple of sections overlap.
Suggestions
Trim or move 'Locating Elements' and 'Assertions Reference' into a references file (or cut to only the project-specific rules, e.g. 'never waitForTimeout, always auto-retrying expect, getByTestId only as last resort') — Claude already knows the Playwright API.
Consolidate the 'Run Directly or from /use-case-test' rules with Workflow step 8, which repeats the failure-handling mechanics, to remove the overlap.
Add a one-line pointer to references/layer.md (or note that it is a tooling-discovered convention file) so every bundle file is navigable from the body.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | Mostly efficient and dense with project-specific conventions, but 'Locating Elements' and 'Assertions Reference' re-teach standard Playwright API knowledge Claude already has (commented getByRole basics, expect(...).toBeVisible() examples), and the 'Run Directly' section overlaps Workflow step 8. Not score 4 because the over-explanation is more than minor; not score 2 because most content (test-data policy, naming conventions, markers, scope rules) is genuinely non-obvious. | 3 / 5 |
Actionability | Fully executable throughout: install commands, a complete playwright.config.ts, exact file-naming patterns, run commands (npx playwright test --grep "@UC-010"), a section-to-code mapping table for test case documents, external templates, and a concrete DO NOT list covering common cases. | 5 / 5 |
Workflow Clarity | A clearly sequenced 10-step workflow with an explicit validation and feedback loop: run the tests, and on failure apply a concrete debugging checklist (frontend/backend up, proxy, seed data, exact accessible name, assert heading first) with --headed --debug or the MCP server, then rerun; the final report step demands step, expected, and actual values. | 5 / 5 |
Progressive Disclosure | Templates and marker rows are appropriately split into one-level-deep, clearly linked references files (references/use-case.spec.ts.md, references/test-case.spec.ts.md, references/markers.md — all present). Minor gaps: references/layer.md is never signposted from the body (it is found by external tooling via glob), and ~40 lines of standard Playwright API reference sit inline in 'Locating Elements'/'Assertions Reference' where a reference file would do. Not score 5 because of these minor organization gaps; not score 3 because the structure, signaling, and split decisions are otherwise sound. | 4 / 5 |
Total | 17 / 20 Passed |