Content
71%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 highly actionable, well-structured audit rubric with executable commands, precise thresholds, and a complete output template — its strongest dimension by far. The weaknesses are moderate verbosity from explaining concepts Claude already knows and a monolithic single-file layout that inlines template and reference material that belongs in bundle files.
Suggestions
Trim explanations of concepts Claude already knows (what CLS/INP measure, what RFC 8288 and IndexNow are, why mobile-first indexing exists) and keep only the skill-specific thresholds, nuances, and decisions.
Move the full GEO-TECHNICAL-AUDIT.md output template and the AI-crawler user-agent table into references/ files (e.g., references/output-template.md, references/ai-crawlers.md), keeping a short pointer and the 2-3 most decision-critical rows inline.
Add error-handling guidance for the fetch steps (non-200 responses, unreachable robots.txt/sitemap, malformed XML) so the workflow has explicit recovery checkpoints rather than handling them only in Category 9.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The body is mostly efficient — checklists, tables, thresholds, and point values rather than prose — but it includes several explanations of concepts Claude already knows, e.g., "INP ... Measures ALL interactions, not just first", "CLS ... Measures visual stability — unexpected layout movements", "IndexNow is an open protocol that allows websites to notify search engines instantly", and "RFC 8288 (Web Linking) defines the HTTP Link: response header". This fits the 'mostly efficient but includes some unnecessary explanation' anchor; it is not a 2 because the padding is incidental rather than structural, and not a 4 because the known-concept explanations recur across multiple categories. | 3 / 5 |
Actionability | Guidance is fully executable throughout: exact commands ("curl -o /dev/null -s -w 'TTFB: %{time_starttransfer}s\n' [URL]", "curl -s [URL]"), precise thresholds (LCP < 2.5s, TTFB < 800ms, page weight < 2MB, 48x48px tap targets), a complete AI-crawler user-agent table, per-category point rubrics, and a copy-paste-ready output template. This matches the 'fully executable; specific examples cover the common cases' anchor. | 5 / 5 |
Workflow Clarity | "How to Use This Skill" gives a clear 5-step sequence (collect URL, fetch, run categories, score, generate report), reinforced by explicit interpretation bands (90-100 Excellent ... 0-29 Critical) and per-category Pass/Warn/Fail thresholds that act as checkpoints. It is a 4 rather than 5 because error-recovery loops are missing for the fetch steps (what to do when a fetch fails, a page returns non-200, or the sitemap is malformed — only Category 9 specifies skip-and-note behavior); it is not a 3 because the sequence, scoring checkpoints, and output validation via status thresholds are all present. | 4 / 5 |
Progressive Disclosure | The single SKILL.md has no bundle files (no references/, scripts/, or assets/ exist), so all 533 lines — including the full output template, the 11-row AI-crawler table, and two full pages of boilerplate per-scenario markdown for the report — are inlined. Section headers provide reasonable navigation, but content that clearly belongs in separate reference files (e.g., OUTPUT-TEMPLATE.md, AI-CRAWLERS.md) sits inline, matching the 'some structure but content that should be separate is inline' anchor; it is not a 2 because the document is well-sectioned and navigable rather than an unstructured wall. | 3 / 5 |
Total | 15 / 20 Passed |