Content
67%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 well-structured instructional skill: concrete numbers, named tools and APIs, a sequenced workflow with re-measurement checkpoints, and clean one-level references. Its main weakness is token efficiency — much of the Core Web Vitals and best-practices material restates standard knowledge Claude already has and overlaps the referenced playbook.
Suggestions
Move the per-metric 'What it measures / Common causes / Common fixes' detail into references/optimization-playbook.md and keep only the metric names, target thresholds, and a pointer in SKILL.md — Claude already knows what LCP, INP, and CLS measure.
Trim the 'Beyond Core Web Vitals' best-practice lists (images, fonts, third parties) to the non-obvious specifics (e.g., the 170KB/100KB budget numbers, the 60-80 percent image-weight figure) and let references/optimization-checklist.md carry the rest.
Add one or two executable commands (e.g., a Lighthouse CLI invocation for the baseline step) so the workflow's first step is copy-paste ready.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dense and free of filler, but large portions restate knowledge Claude already has: "What it measures: Time until the largest visible content element finishes rendering", standard CWV target tiers, and well-known image/font/third-party best practices. This is more than minor over-explanation (not 4), but there is no padded or conversational fluff (not 2) — the material could simply be tightened by moving textbook detail into the referenced files. | 3 / 5 |
Actionability | Concrete, specific guidance throughout: numeric thresholds ("TTFB over 800ms", "Initial JS under 170KB compressed", "under 1500 elements"), named APIs (`scheduler.yield()`, `font-display: optional`, `srcset`, `<link rel="preload" as="font">`), and named tools with dev-tools tabs. Falls short of 5 because nothing is copy-paste executable — e.g., no Lighthouse CLI command or code snippet for any fix — though for an instruction-only skill this is a minor gap, not the incompleteness of a 3. | 4 / 5 |
Workflow Clarity | The 7-step workflow is clearly sequenced with an explicit validation checkpoint ("Re-measure. After each major fix, re-run Lighthouse and check Real User Monitoring") and the failure-patterns section reinforces correct ordering. Not 5 because there is no explicit error-recovery loop for when a fix fails to move the metric, and the checklist lives in a reference rather than the workflow itself. | 4 / 5 |
Progressive Disclosure | Three real, one-level-deep reference files are clearly signaled with per-file purpose descriptions, and SKILL.md carries the framework, workflow, and output format. Not 5 because SKILL.md is heavier than an overview: the inlined per-metric causes/fixes lists substantially duplicate the role of the referenced optimization-playbook.md ("Symptom-to-fix playbook for the common Core Web Vitals problems"), content that could be split out. | 4 / 5 |
Total | 15 / 20 Passed |