Content
88%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 body: exact MCP tool calls, a validated 5-phase workflow with troubleshooting loops and skip conditions, and clean tabular organization. The main improvement areas are trimming knowledge Claude already has (inlined CWV thresholds, framework config filenames) and moving the Phase 5 codebase-analysis detail into a reference file.
Suggestions
Move the Phase 5 codebase-analysis material (framework detection table, tree-shaking, polyfills, compression checks) into a references/codebase-analysis.md and link to it one level deep, keeping SKILL.md as a lean overview.
Replace or prune the inlined Core Web Vitals threshold table since the skill itself directs retrieval from web.dev/Lighthouse docs for current thresholds — keep only the retrieval pointers.
Trim the framework config-filename table (webpack.config.js, vite.config.ts, etc.), which is knowledge Claude already has, down to a one-line instruction to detect the bundler from config files.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is table-driven and largely free of fluff — tool-call quick reference, insight names, and a 5-phase checklist with no filler prose. Minor over-explanation of things Claude already knows: the inlined Core Web Vitals threshold table (which the skill itself says should come from retrieval) and the framework config-filename table ("Webpack: webpack.config.js", etc.), both of which could be trimmed. It sits above the score-3 anchor because the padding is minor, not 'some unnecessary explanation'. | 4 / 5 |
Actionability | Fully concrete and executable: exact tool signatures like `performance_start_trace(autoStop: true, reload: true)`, `performance_analyze_insight(insightSetId: "<id-from-trace>", insightName: "LCPBreakdown")`, `list_network_requests(resourceTypes: ["Script", "Stylesheet", ...])`, plus real insight names (LCPBreakdown, CLSCulprits) and an MCP config snippet. The Quick Reference table makes the common cases copy-paste ready, matching the score-5 anchor. | 5 / 5 |
Workflow Clarity | A clearly sequenced 5-phase workflow with a copyable progress checklist, an explicit pre-flight validation step ("Run this before starting... If unavailable, STOP"), troubleshooting feedback loops in Phase 1 and Phase 2 ("If trace returns empty or fails, verify the page loaded...", "If an insight name doesn't work, check the insightSetId..."), and explicit skip conditions (Phase 5 "Skip if auditing a third-party site"). This matches the anchor requiring explicit validation steps and error-recovery loops. | 5 / 5 |
Progressive Disclosure | No bundle files exist (references/, scripts/, assets/ are absent), so all content is inlined in a ~200-line SKILL.md. Section headers are clear and well-organized and external retrieval sources are cleanly listed, but content like the framework-detection table and codebase-analysis subsections (polyfills, tree-shaking, compression) would fit naturally in a one-level-deep reference file. Good structure with minor organization gaps — the score-4 anchor. | 4 / 5 |
Total | 18 / 20 Passed |