Content
85%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 high-quality, highly actionable skill body with concrete code, exact tooling, numeric WCAG thresholds, and an audit workflow that includes an explicit feedback loop and completion-evidence checklist. Its main weakness is progressive disclosure: it points to many policy/template/script files that are absent from this bundle, leaving those references unverifiable.
Suggestions
Ship the referenced files (at minimum `scripts/check-contrast.mjs` and `templates/accessibility-check.md`) inside the skill's scripts/ and references/ directories, or remove/inline the references — the contrast-check command is presented as the canonical verification gate but has no backing file in the bundle.
Convert bare path mentions like `policies/handoffs.md` and `templates/accessibility-check.md` into clearly signaled, one-level-deep links (e.g., "See [templates/accessibility-check.md](templates/accessibility-check.md)") so navigation to detail files is explicit rather than buried in prose.
Tighten the Handoff section: the two long paragraphs about skills 56 and 57 restate overlapping scope boundaries and could be reduced to two lines each, trimming the main token overhead in an otherwise lean body.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and token-efficient — numeric thresholds ("4.5:1 texto normal, 3:1 texto grande"), tool names with exact packages, and a prioritized "criterios que mais quebram" checklist rather than a full WCAG dump — but the governance section and the two long handoff paragraphs (skills 56/57) add prose that could be trimmed, landing it just below the every-token-earns-its-place anchor. | 4 / 5 |
Actionability | Guidance is copy-paste ready and covers common cases: complete ERRADO/CERTO HTML/CSS snippets (aria-label on icon buttons, aria-describedby error linking, prefers-reduced-motion media query), exact tool invocations ("@axe-core/playwright", "jest-axe", "NVDA (Win, gratis) ou VoiceOver (Mac, Cmd+F5)"), and a runnable verification command (`node scripts/check-contrast.mjs <css>`). | 5 / 5 |
Workflow Clarity | The "Checkpoint" paragraph gives an explicit sequence — run the 4 test categories in table order, turn each finding into a concrete fix, apply, then re-run only the failed category — with a genuine feedback loop that even distinguishes symptom from cause ("fix cobriu o sintoma... mas nao a causa"), reinforced by the "Evidencia de Conclusao" validation checklist. | 5 / 5 |
Progressive Disclosure | Sections are well organized and the inline prioritized checklist is appropriately sized for SKILL.md, but the body references a dozen files — `GLOBAL.md`, `policies/execution.md`, `policies/handoffs.md`, `scripts/check-contrast.mjs`, `templates/accessibility-check.md`, `memory/constitution.md` — and none of them exist in the bundle (no references/, scripts/, or assets/ directories are present), so the references are neither verifiable nor clearly signaled as one-level-deep navigations. | 3 / 5 |
Total | 17 / 20 Passed |