Content
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
A well-structured, actionable skill body with executable code across Playwright/Cypress/raw API, a clear gated workflow with a feedback loop, and correct one-level-deep reference splitting. The only weakness is mild verbosity and duplication that could be tightened without losing information.
Suggestions
Remove the duplicated 'persist the raw accessibilityScanResults JSON' guidance - it appears in both 'How to use' step 5 and the 'CI integration' closing note; consolidate to one location.
Tighten the Overview opening: 'axe-core is "an accessibility testing engine for websites and other HTML-based user interfaces" maintained by Deque' defines a tool Claude already knows - lead instead with the non-obvious 57% auto-detection stat and the integration shape.
Reconcile the frontmatter description, which promises to 'emit JUnit-shaped output for CI gating', with the body, which uses a11y-violation-gate instead - either add a brief JUnit emit example or adjust the description to match.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly lean and code-forward, but it includes unnecessary explanation ('axe-core is "an accessibility testing engine for websites and other HTML-based user interfaces" maintained by Deque' defines a tool Claude already knows) and duplications - 'persist the raw accessibilityScanResults JSON' appears in both 'How to use' step 5 and the 'CI integration' closing note, and the framework-wrapper guidance repeats the Install table - so it sits at the score-2 'could be tightened' anchor rather than 'every token earns its place'. | 2 / 3 |
Actionability | It provides copy-paste-ready, executable guidance throughout - the raw `axe.run()` JS example, the Playwright `AxeBuilder` TypeScript snippet, the Cypress `cypress-axe` commands, `.withTags(...)`, `.disableRules(...)`, `.exclude(...)`, the CI workflow YAML, and `jq` triage - matching the score-3 'fully executable code/commands' anchor. | 3 / 3 |
Workflow Clarity | The 7-step 'How to use' sequence (install -> add a state-driven scan test -> pin conformance with withTags -> suppress narrowly -> persist JSON -> feed to a11y-violation-gate -> route incomplete to manual review) is clearly ordered, and the Worked example demonstrates an explicit validate/fix/retry feedback loop (gate fails the PR -> fix the button colour -> next run reports violations: []), matching the score-3 anchor; this is a scan/CI task, not a destructive operation, so the destructive-cap does not apply. | 3 / 3 |
Progressive Disclosure | SKILL.md stays an overview of core scans while pushing the heavy reference material (results arrays, violation field tables, tag sets, disable/exclude, jq triage) one level deep into the real, clearly-signaled [references/results-and-config.md](references/results-and-config.md) file that exists in the bundle, matching the score-3 'clear overview with well-signaled one-level-deep references' anchor. | 3 / 3 |
Total | 11 / 12 Passed |