Content
82%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, operationally dense skill: executable config, commands, and CI workflow throughout, with genuinely non-obvious guidance (INP timespan limitation, TBT proxy, anti-patterns) and a properly delegated one-level-deep reference file. Remaining gaps are minor: trimming background explanation, an explicit pre-CI validation step, and splitting a few inline sections into references.
Suggestions
Trim background Claude already knows: replace the Core Web Vitals definitions table and the mobile-slower rationale with one-line pointers to the thresholds, keeping only the operative facts (values, 75th-percentile standard, INP-replaces-FID).
Add an explicit validation checkpoint before CI wiring, e.g. "Run `lhci autorun` locally first; only add the workflow once it exits 0, and start assertions at warn level".
Consider moving the mobile-vs-desktop dual-config details and the lab-vs-field table into references/budgets.md (or a companion reference) to shorten the main file and keep SKILL.md as an overview.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is dominated by genuinely non-obvious operational knowledge (INP is timespan-only in Lighthouse so gate on TBT; warn-then-promote rollout; a dense anti-patterns table), but a few passages explain things Claude already knows — the Core Web Vitals definitions table and "mobile is consistently slower in practice - the same JS bundle runs on a less-powerful CPU". Matches the 4 anchor (efficient with minor over-explanation that could be trimmed), not 5. | 4 / 5 |
Actionability | Fully executable throughout: a complete `.lighthouserc.js` with real assertion syntax, `lhci autorun` plus per-phase and single-URL invocations, a copy-paste GitHub Actions workflow with artifact upload, and dual mobile/desktop run commands. Matches the 5 anchor (copy-paste ready, covering the common cases). | 5 / 5 |
Workflow Clarity | Install → Configure → Run → CI integration is a clear, concrete sequence, and the anti-patterns table supplies a rollout checkpoint ("Start with warn for everything; promote to error once green for 2 weeks"). However, there is no explicit validation step such as verifying the first local run passes before enabling error-level gating in CI, matching the 4 anchor (most checkpoints present, minor validation gaps) rather than 5's explicit validate-and-recover loops. No destructive or batch operations, so the workflow cap does not apply. | 4 / 5 |
Progressive Disclosure | The bundle is well-structured: `references/budgets.md` exists, is clearly signaled in a dedicated "Authoring budgets" section with "Start there when..." guidance, and is one level deep (verified a leaf file with no nested references). But at 267 lines some inline sections (mobile-vs-desktop, lab-vs-field, anti-patterns) could arguably be split into references, and the References section mixes link definitions with sibling-skill pointers — matching the 4 anchor (good structure, minor organization gaps) rather than 5. | 4 / 5 |
Total | 17 / 20 Passed |