Content
35%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This skill reads more like a comprehensive web development reference guide than a focused skill for Claude. It extensively explains concepts Claude already knows (Lighthouse categories, WCAG criteria, SEO basics) without providing novel tooling, executable commands, or specific workflows. The output format template and severity classification are useful, but the bulk of the content is redundant general knowledge that wastes token budget.
Suggestions
Remove the detailed explanations of each audit category (Performance, Accessibility, SEO, Best Practices) since Claude already knows these — instead, provide only the specific thresholds, output format, and severity classification that represent novel guidance.
Add concrete, executable instructions for how Claude should actually perform the audit (e.g., what tools to invoke, what files to inspect, what commands to run) rather than listing what good web practices are.
Add validation/feedback loops to the workflow: e.g., after identifying issues, verify fixes against specific criteria before marking them resolved.
Leverage the referenced specialized skills more aggressively — keep this SKILL.md as a brief orchestration overview (~30-40 lines) that delegates depth to the linked skills.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | This skill is extremely verbose at ~150+ lines, largely restating well-known web development best practices that Claude already knows (what LCP is, what alt text is, what HTTPS is, etc.). The vast majority of content is a textbook-style enumeration of Lighthouse audit categories rather than novel, Claude-specific guidance. | 1 / 3 |
Actionability | The skill provides structured checklists and a clear output format template, which is somewhat actionable. However, it lacks any executable code, concrete commands (e.g., how to actually run Lighthouse), or specific tooling instructions. The guidance is descriptive ('optimize images', 'remove unused code') rather than executable. | 2 / 3 |
Workflow Clarity | The four-step 'How it works' section provides a basic sequence, and the output format gives structure to results. However, there are no validation checkpoints, no feedback loops for error recovery, and no clear decision points for when to escalate or re-audit after fixes. | 2 / 3 |
Progressive Disclosure | References to specialized skills (performance, accessibility, SEO, best-practices) are provided at the bottom, which is good. However, the SKILL.md itself contains extensive inline content that duplicates what those referenced skills presumably cover in detail, rather than serving as a concise overview pointing to them. No bundle files exist to verify the references resolve. | 2 / 3 |
Total | 7 / 12 Passed |